fitting smooth in time prognostic risk functions via
play

Fitting smooth-in-time prognostic risk functions via logistic - PowerPoint PPT Presentation

Oct 26, 2023 •385 likes •2.04k views

Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary Fitting smooth-in-time prognostic risk functions via logistic regression James A. Hanley 1 Olli S. Miettinen 1 1 Department of

  1. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  2. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  3. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  4. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  5. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  6. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  7. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  8. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  9. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FORM h ( x , t ) = e g ( x , t ,β ) log { h ( x , t ) } = g ( x , t , β ) ⇐ ⇒ • x is a realization of the covariate vector X , representing the patient profile P , and possible intervention I . • β : a vector of parameters with unknown values, • g () includes constant 1, variates for P , I and t ; • g () can have product terms involving P , I , and t . • g () must be ‘linear’ in parameters, in ‘ linear model’ sense. ————– • ‘proportional hazards’ if no product terms involving t & I • If t is represented by a linear term (so that ‘time to event’ ∼ Gompertz ), then � CI p , i ( t ) has a closed smooth form. • If t is replaced by log t , then ‘time to event’ ∼ Weibull .

  10. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FITTING • Unable to find a ready-to-use ML procedure within the common statistical packages. • Likelihood becomes quite involved even if no censored observations. • Albertsen & Hanley(’98); Efron(’88, ’02); Carstensen(’00): - divide ‘survival time’ of each subject into time-slices; - treat # of events in each ∼ Binomial / Poisson.

  11. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FITTING • Unable to find a ready-to-use ML procedure within the common statistical packages. • Likelihood becomes quite involved even if no censored observations. • Albertsen & Hanley(’98); Efron(’88, ’02); Carstensen(’00): - divide ‘survival time’ of each subject into time-slices; - treat # of events in each ∼ Binomial / Poisson.

  12. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FITTING • Unable to find a ready-to-use ML procedure within the common statistical packages. • Likelihood becomes quite involved even if no censored observations. • Albertsen & Hanley(’98); Efron(’88, ’02); Carstensen(’00): - divide ‘survival time’ of each subject into time-slices; - treat # of events in each ∼ Binomial / Poisson.

  13. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FULLY-PARAMETRIC MODEL: FITTING • Unable to find a ready-to-use ML procedure within the common statistical packages. • Likelihood becomes quite involved even if no censored observations. • Albertsen & Hanley(’98); Efron(’88, ’02); Carstensen(’00): - divide ‘survival time’ of each subject into time-slices; - treat # of events in each ∼ Binomial / Poisson.

  14. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  15. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  16. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  17. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  18. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  19. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  20. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  21. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTING: OUR APPROACH • An extension of the method of Mantel (1973) to binary outcomes that deals with time dimension. • Mantel’s problem: • ( c =) 165 ‘cases’ of Y = 1 , • 4000 instances of Y = 0 . • Associated regressor vector X for each of the 4165 • A logistic model for Prob ( Y = 1 | X ) • A computer with limited capacity.

  22. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary MANTEL ’S SOLUTION • Form a reduced dataset containing... • All c instances (cases) of Y = 1 • Random sample of the Y = 0 observations • Fit the same logistic model to this reduced dataset. “Such sampling will tend to leave the dependence of the log odds on the variables unaffected except for an additive constant.” Anderson (Biometrika, 1972) had noted this too. • Outcome(Choice)-based sampling common in Epi, Marketing, etc...

  23. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary MANTEL ’S SOLUTION • Form a reduced dataset containing... • All c instances (cases) of Y = 1 • Random sample of the Y = 0 observations • Fit the same logistic model to this reduced dataset. “Such sampling will tend to leave the dependence of the log odds on the variables unaffected except for an additive constant.” Anderson (Biometrika, 1972) had noted this too. • Outcome(Choice)-based sampling common in Epi, Marketing, etc...

  24. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary MANTEL ’S SOLUTION • Form a reduced dataset containing... • All c instances (cases) of Y = 1 • Random sample of the Y = 0 observations • Fit the same logistic model to this reduced dataset. “Such sampling will tend to leave the dependence of the log odds on the variables unaffected except for an additive constant.” Anderson (Biometrika, 1972) had noted this too. • Outcome(Choice)-based sampling common in Epi, Marketing, etc...

  25. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary MANTEL ’S SOLUTION • Form a reduced dataset containing... • All c instances (cases) of Y = 1 • Random sample of the Y = 0 observations • Fit the same logistic model to this reduced dataset. “Such sampling will tend to leave the dependence of the log odds on the variables unaffected except for an additive constant.” Anderson (Biometrika, 1972) had noted this too. • Outcome(Choice)-based sampling common in Epi, Marketing, etc...

  26. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary MANTEL ’S SOLUTION • Form a reduced dataset containing... • All c instances (cases) of Y = 1 • Random sample of the Y = 0 observations • Fit the same logistic model to this reduced dataset. “Such sampling will tend to leave the dependence of the log odds on the variables unaffected except for an additive constant.” Anderson (Biometrika, 1972) had noted this too. • Outcome(Choice)-based sampling common in Epi, Marketing, etc...

  27. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary MANTEL ’S SOLUTION • Form a reduced dataset containing... • All c instances (cases) of Y = 1 • Random sample of the Y = 0 observations • Fit the same logistic model to this reduced dataset. “Such sampling will tend to leave the dependence of the log odds on the variables unaffected except for an additive constant.” Anderson (Biometrika, 1972) had noted this too. • Outcome(Choice)-based sampling common in Epi, Marketing, etc...

  28. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary MANTEL ’S SOLUTION • Form a reduced dataset containing... • All c instances (cases) of Y = 1 • Random sample of the Y = 0 observations • Fit the same logistic model to this reduced dataset. “Such sampling will tend to leave the dependence of the log odds on the variables unaffected except for an additive constant.” Anderson (Biometrika, 1972) had noted this too. • Outcome(Choice)-based sampling common in Epi, Marketing, etc...

  29. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary DATA TO EXPLAIN OUR APPROACH Systolic Hypertension in Elderly Program (SHEP) .......................... SHEP Cooperative Research Group (1991). .......................... Journal of American Medical Association 265, 3255-3264. • 4,701 persons with complete data on P = { age, sex, race, and systolic blood pressure} and I = { active/placebo}. • Study base of B = 20 , 894 person-years of follow-up; c = 263 events ("cases") of stroke identified.

  30. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary DATA TO EXPLAIN OUR APPROACH Systolic Hypertension in Elderly Program (SHEP) .......................... SHEP Cooperative Research Group (1991). .......................... Journal of American Medical Association 265, 3255-3264. • 4,701 persons with complete data on P = { age, sex, race, and systolic blood pressure} and I = { active/placebo}. • Study base of B = 20 , 894 person-years of follow-up; c = 263 events ("cases") of stroke identified.

  31. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary DATA TO EXPLAIN OUR APPROACH Systolic Hypertension in Elderly Program (SHEP) .......................... SHEP Cooperative Research Group (1991). .......................... Journal of American Medical Association 265, 3255-3264. • 4,701 persons with complete data on P = { age, sex, race, and systolic blood pressure} and I = { active/placebo}. • Study base of B = 20 , 894 person-years of follow-up; c = 263 events ("cases") of stroke identified.

  32. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary DATA TO EXPLAIN OUR APPROACH Systolic Hypertension in Elderly Program (SHEP) .......................... SHEP Cooperative Research Group (1991). .......................... Journal of American Medical Association 265, 3255-3264. • 4,701 persons with complete data on P = { age, sex, race, and systolic blood pressure} and I = { active/placebo}. • Study base of B = 20 , 894 person-years of follow-up; c = 263 events ("cases") of stroke identified.

  33. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary DATA TO EXPLAIN OUR APPROACH Systolic Hypertension in Elderly Program (SHEP) .......................... SHEP Cooperative Research Group (1991). .......................... Journal of American Medical Association 265, 3255-3264. • 4,701 persons with complete data on P = { age, sex, race, and systolic blood pressure} and I = { active/placebo}. • Study base of B = 20 , 894 person-years of follow-up; c = 263 events ("cases") of stroke identified.

  34. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary STUDY BASE, and the 263 cases 6000 − 20,894 person−years [B=20,894 PY] STUDY BASE − 10,982,000,000 person−minutes (approx) 5000 No. of Persons − infinite number of person−moments ↑ Being Followed c = 263 events (Y=1) ● ● ● ● ● ● ● in this infinite number ● ● ● 4000 ● ● ● ● ● ● ● ● ● ● ● ● ● of person−moments ● ● ● ● ● ● ● ● ● ● ● ● ● Persons ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● infinite number ● ● ● ● ● ● ● ● 3000 ● ● ● ● ● ● ● ● ● ● of person−moments ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● with Y=0 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 2000 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 1000 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0 ● ● ● ● ● 0 1 2 3 4 5 6 7 t: Years since Randomization

  35. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary OUR APPROACH • Base series: representative (unstratified) sample of base. • b : size of base series • B : amount of population-time constituting study base. • B ( x , t ) : population-time element in study base Pr ( Y = 1 | x , t ) Pr ( Y = 0 | x , t ) = h ( x , t ) × B ( x , t ) b × [ B ( x , t ) / B ] = h ( x , t ) × ( B / b ) , • log ( B / b ) is an offset [a regression term with known coefficient of 1] . → logistic model, with t having same status as x , and offset, directly yields � ID x , t = exp { � h ( x , t ) = � g ( x , t ) } .

  36. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary How large should b be on relation to c ? b : no. of instances of Y = 0 ; c : no. of instances of Y = 1 • Mantel (1973)... little to be gained by letting the size of one series, b, become arbitrarily large if the size of the other series, c , must remain fixed. • With 2008 computing, we can use a b / c ratio as high as 100, and thereby extract virtually all of the information in the base.

  37. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary How large should b be on relation to c ? b : no. of instances of Y = 0 ; c : no. of instances of Y = 1 • Mantel (1973)... little to be gained by letting the size of one series, b, become arbitrarily large if the size of the other series, c , must remain fixed. • With 2008 computing, we can use a b / c ratio as high as 100, and thereby extract virtually all of the information in the base.

  38. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary How large should b be on relation to c ? b : no. of instances of Y = 0 ; c : no. of instances of Y = 1 • Mantel (1973)... little to be gained by letting the size of one series, b, become arbitrarily large if the size of the other series, c , must remain fixed. • With 2008 computing, we can use a b / c ratio as high as 100, and thereby extract virtually all of the information in the base.

  39. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary How large should b be on relation to c ? b : no. of instances of Y = 0 ; c : no. of instances of Y = 1 • Mantel (1973)... little to be gained by letting the size of one series, b, become arbitrarily large if the size of the other series, c , must remain fixed. • With 2008 computing, we can use a b / c ratio as high as 100, and thereby extract virtually all of the information in the base.

  40. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary OUR HAZARD MODEL FOR SHEP DATA log [ h ] = Σ β k X k , where X 1 = Age (in yrs) - 60 X 2 = Indicator of male gender X 3 = Indicator of Black race X 4 = Systolic BP (in mmHg) - 140 ...................................................................... X 5 = Indicator of active treatment ...................................................................... X 6 = T ...................................................................... X 7 = X 5 × X 6 . (non-proportional hazards)

  41. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary OUR HAZARD MODEL FOR SHEP DATA log [ h ] = Σ β k X k , where X 1 = Age (in yrs) - 60 X 2 = Indicator of male gender X 3 = Indicator of Black race X 4 = Systolic BP (in mmHg) - 140 ...................................................................... X 5 = Indicator of active treatment ...................................................................... X 6 = T ...................................................................... X 7 = X 5 × X 6 . (non-proportional hazards)

  42. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  43. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  44. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  45. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  46. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  47. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  48. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  49. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary PARAMETER ESTIMATION • Formed person-moments dataset pertaining to: • case series of size c = 263 ( Y = 1) and • (randomly-selected) base series of size b = 26 , 300 ( Y = 0 ) . • Each of 26,563 rows contained realizations of • X 1 , . . . , X 7 • Y • offset = log ( 20 , 894 / 26 , 300 ) . • Logistic model fitted to data in the two series.

  50. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary DATASET: c = 263 ; b = 10 × 263 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4000 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 3000 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● Persons ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 2000 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 1000 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0 1 2 3 4 5 6 Time

  51. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTED VALUES Proposed Cox logistic regression regression β age − 60 0.041 0.041 0.041 β I male 0.257 0.258 0.259 β I black 0.302 0.301 0.303 β SBP − 140 0.017 0.017 0.017 .................... β I Active treatment -0.200 -0.435 -0.435 .................... β 0 -5.390 -5.295 β t -0.014 -0.057 β t × I Active treatment -0.107 • Fitted logistic function represents log [ h x ( t )] • → cumulative hazard H X ( t ) , and, thus, X -specific risk.

  52. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTED VALUES Proposed Cox logistic regression regression β age − 60 0.041 0.041 0.041 β I male 0.257 0.258 0.259 β I black 0.302 0.301 0.303 β SBP − 140 0.017 0.017 0.017 .................... β I Active treatment -0.200 -0.435 -0.435 .................... β 0 -5.390 -5.295 β t -0.014 -0.057 β t × I Active treatment -0.107 • Fitted logistic function represents log [ h x ( t )] • → cumulative hazard H X ( t ) , and, thus, X -specific risk.

  53. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTED VALUES Proposed Cox logistic regression regression β age − 60 0.041 0.041 0.041 β I male 0.257 0.258 0.259 β I black 0.302 0.301 0.303 β SBP − 140 0.017 0.017 0.017 .................... β I Active treatment -0.200 -0.435 -0.435 .................... β 0 -5.390 -5.295 β t -0.014 -0.057 β t × I Active treatment -0.107 • Fitted logistic function represents log [ h x ( t )] • → cumulative hazard H X ( t ) , and, thus, X -specific risk.

  54. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary FITTED VALUES Proposed Cox logistic regression regression β age − 60 0.041 0.041 0.041 β I male 0.257 0.258 0.259 β I black 0.302 0.301 0.303 β SBP − 140 0.017 0.017 0.017 .................... β I Active treatment -0.200 -0.435 -0.435 .................... β 0 -5.390 -5.295 β t -0.014 -0.057 β t × I Active treatment -0.107 • Fitted logistic function represents log [ h x ( t )] • → cumulative hazard H X ( t ) , and, thus, X -specific risk.

  55. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary ESTIMATED 5-YEAR RISK OF STROKE Risk I h ( t ) H ( 5 ) CI ( 5 ) ∆ � 5 [ 1 − e − H ( 5 ) ] [ ID(t) ] [ 0 h x ( t ) dt ] e − 4 . 86 − 0 . 014 t Low 0 0.037 0.036 e − 5 . 06 − 0 . 124 t 1 0.024 0.024 1.2% High 0 0.16 1 0.10 6% Overall 0 0.076 1 0.049 2.7% Low: 65 year old white female with a SBP of 160 mmHg. High: 80 year old black male with a SBP of 180 mmHg

  56. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary ESTIMATED 5-YEAR RISK OF STROKE Risk I h ( t ) H ( 5 ) CI ( 5 ) ∆ � 5 [ 1 − e − H ( 5 ) ] [ ID(t) ] [ 0 h x ( t ) dt ] e − 4 . 86 − 0 . 014 t Low 0 0.037 0.036 e − 5 . 06 − 0 . 124 t 1 0.024 0.024 1.2% High 0 0.16 1 0.10 6% Overall 0 0.076 1 0.049 2.7% Low: 65 year old white female with a SBP of 160 mmHg. High: 80 year old black male with a SBP of 180 mmHg

  57. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary ESTIMATED 5-YEAR RISK OF STROKE Risk I h ( t ) H ( 5 ) CI ( 5 ) ∆ � 5 [ 1 − e − H ( 5 ) ] [ ID(t) ] [ 0 h x ( t ) dt ] e − 4 . 86 − 0 . 014 t Low 0 0.037 0.036 e − 5 . 06 − 0 . 124 t 1 0.024 0.024 1.2% High 0 0.16 1 0.10 6% Overall 0 0.076 1 0.049 2.7% Low: 65 year old white female with a SBP of 160 mmHg. High: 80 year old black male with a SBP of 180 mmHg

  58. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary ESTIMATED 5-YEAR RISK OF STROKE Risk I h ( t ) H ( 5 ) CI ( 5 ) ∆ � 5 [ 1 − e − H ( 5 ) ] [ ID(t) ] [ 0 h x ( t ) dt ] e − 4 . 86 − 0 . 014 t Low 0 0.037 0.036 e − 5 . 06 − 0 . 124 t 1 0.024 0.024 1.2% High 0 0.16 1 0.10 6% Overall 0 0.076 1 0.049 2.7% Low: 65 year old white female with a SBP of 160 mmHg. High: 80 year old black male with a SBP of 180 mmHg

  59. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary ESTIMATED 5-YEAR RISK OF STROKE Risk I h ( t ) H ( 5 ) CI ( 5 ) ∆ � 5 [ 1 − e − H ( 5 ) ] [ ID(t) ] [ 0 h x ( t ) dt ] e − 4 . 86 − 0 . 014 t Low 0 0.037 0.036 e − 5 . 06 − 0 . 124 t 1 0.024 0.024 1.2% High 0 0.16 1 0.10 6% Overall 0 0.076 1 0.049 2.7% Low: 65 year old white female with a SBP of 160 mmHg. High: 80 year old black male with a SBP of 180 mmHg

  60. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary ESTIMATED 5-YEAR RISK OF STROKE Risk I h ( t ) H ( 5 ) CI ( 5 ) ∆ � 5 [ 1 − e − H ( 5 ) ] [ ID(t) ] [ 0 h x ( t ) dt ] e − 4 . 86 − 0 . 014 t Low 0 0.037 0.036 e − 5 . 06 − 0 . 124 t 1 0.024 0.024 1.2% High 0 0.16 1 0.10 6% Overall 0 0.076 1 0.049 2.7% Low: 65 year old white female with a SBP of 160 mmHg. High: 80 year old black male with a SBP of 180 mmHg

  61. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary (a.0): 80 year old black male, SBP=180 15 semi−parametric (Cox) (a.1) proposed Cumulative incidence (%) 10 5 (b.0) (b.1): 65 year old white female, SBP=160 0 0 1 2 3 4 5 Prospective time (years)

  62. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 0 1 2 3 4 5 6 7 8 9 10 Points Age 60 65 70 75 80 85 90 95 100 1 Male 0 1 Black 0 SBP 155 165 175 185 195 205 215 0 I 1 t 6 0 I.t 6 5 4 3 2 1 0 Total Points 0 2 4 6 8 10 12 14 16 18 20 22 Linear Predictor −6 −5.5 −5 −4.5 −4 −3.5 −3 −2.5 5−year Risk (%) if not treated 3 4 5 6 7 8 9 12 15 18 5−year Risk (%) if treated 2 3 4 5 6 7 8 9 12 15 18

  63. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 1. FEATURES • Smooth-in- t h ( t ) —and CI’s– not new; fitting procedure is. • Keys: 1. representative sampling of the base; 2. offset. • b / c =100 feasible and adequate.

  64. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 1. FEATURES • Smooth-in- t h ( t ) —and CI’s– not new; fitting procedure is. • Keys: 1. representative sampling of the base; 2. offset. • b / c =100 feasible and adequate.

  65. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 1. FEATURES • Smooth-in- t h ( t ) —and CI’s– not new; fitting procedure is. • Keys: 1. representative sampling of the base; 2. offset. • b / c =100 feasible and adequate.

  66. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 1. FEATURES • Smooth-in- t h ( t ) —and CI’s– not new; fitting procedure is. • Keys: 1. representative sampling of the base; 2. offset. • b / c =100 feasible and adequate.

  67. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 2. MODELLING POSSIBILITIES Log-linear modelling for h x ( t ) via logistic regression ... • Standard methods to assess model fit. • Wide range of functional forms for the t -dimension of h x ( t ) . • Effortless handling of censored data. • Flexibility in modeling non-proportionality over t . • Splines for h ( t ) rather than hr ( t ) .

  68. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 2. MODELLING POSSIBILITIES Log-linear modelling for h x ( t ) via logistic regression ... • Standard methods to assess model fit. • Wide range of functional forms for the t -dimension of h x ( t ) . • Effortless handling of censored data. • Flexibility in modeling non-proportionality over t . • Splines for h ( t ) rather than hr ( t ) .

  69. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 2. MODELLING POSSIBILITIES Log-linear modelling for h x ( t ) via logistic regression ... • Standard methods to assess model fit. • Wide range of functional forms for the t -dimension of h x ( t ) . • Effortless handling of censored data. • Flexibility in modeling non-proportionality over t . • Splines for h ( t ) rather than hr ( t ) .

  70. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 2. MODELLING POSSIBILITIES Log-linear modelling for h x ( t ) via logistic regression ... • Standard methods to assess model fit. • Wide range of functional forms for the t -dimension of h x ( t ) . • Effortless handling of censored data. • Flexibility in modeling non-proportionality over t . • Splines for h ( t ) rather than hr ( t ) .

  71. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 2. MODELLING POSSIBILITIES Log-linear modelling for h x ( t ) via logistic regression ... • Standard methods to assess model fit. • Wide range of functional forms for the t -dimension of h x ( t ) . • Effortless handling of censored data. • Flexibility in modeling non-proportionality over t . • Splines for h ( t ) rather than hr ( t ) .

  72. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 2. MODELLING POSSIBILITIES Log-linear modelling for h x ( t ) via logistic regression ... • Standard methods to assess model fit. • Wide range of functional forms for the t -dimension of h x ( t ) . • Effortless handling of censored data. • Flexibility in modeling non-proportionality over t . • Splines for h ( t ) rather than hr ( t ) .

  73. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 2. MODELLING POSSIBILITIES Log-linear modelling for h x ( t ) via logistic regression ... • Standard methods to assess model fit. • Wide range of functional forms for the t -dimension of h x ( t ) . • Effortless handling of censored data. • Flexibility in modeling non-proportionality over t . • Splines for h ( t ) rather than hr ( t ) .

  74. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 3. CLINICAL POSSIBILITIES / DESIDERATA • PDAs (personal digital assistants) → online information. • Profile-specific risk estimates for various interventions. • Already, online calculators: risk of MI, Breast/Lung Cancer; probability of extra-organ spread of cancer. • RCT reports should contain: suitably designed risk function, fitted parameters of h x ( t ) , and risk function. • (Offline:) risk scores → risks via nomogram/table.

  75. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 3. CLINICAL POSSIBILITIES / DESIDERATA • PDAs (personal digital assistants) → online information. • Profile-specific risk estimates for various interventions. • Already, online calculators: risk of MI, Breast/Lung Cancer; probability of extra-organ spread of cancer. • RCT reports should contain: suitably designed risk function, fitted parameters of h x ( t ) , and risk function. • (Offline:) risk scores → risks via nomogram/table.

  76. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 3. CLINICAL POSSIBILITIES / DESIDERATA • PDAs (personal digital assistants) → online information. • Profile-specific risk estimates for various interventions. • Already, online calculators: risk of MI, Breast/Lung Cancer; probability of extra-organ spread of cancer. • RCT reports should contain: suitably designed risk function, fitted parameters of h x ( t ) , and risk function. • (Offline:) risk scores → risks via nomogram/table.

  77. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 3. CLINICAL POSSIBILITIES / DESIDERATA • PDAs (personal digital assistants) → online information. • Profile-specific risk estimates for various interventions. • Already, online calculators: risk of MI, Breast/Lung Cancer; probability of extra-organ spread of cancer. • RCT reports should contain: suitably designed risk function, fitted parameters of h x ( t ) , and risk function. • (Offline:) risk scores → risks via nomogram/table.

  78. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 3. CLINICAL POSSIBILITIES / DESIDERATA • PDAs (personal digital assistants) → online information. • Profile-specific risk estimates for various interventions. • Already, online calculators: risk of MI, Breast/Lung Cancer; probability of extra-organ spread of cancer. • RCT reports should contain: suitably designed risk function, fitted parameters of h x ( t ) , and risk function. • (Offline:) risk scores → risks via nomogram/table.

  79. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 3. CLINICAL POSSIBILITIES / DESIDERATA • PDAs (personal digital assistants) → online information. • Profile-specific risk estimates for various interventions. • Already, online calculators: risk of MI, Breast/Lung Cancer; probability of extra-organ spread of cancer. • RCT reports should contain: suitably designed risk function, fitted parameters of h x ( t ) , and risk function. • (Offline:) risk scores → risks via nomogram/table.

  80. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 4. SUMMARY • Profile-specific risk (CI) functions are important. • Two paths to CI, via... • Steps-in-time S 0 ( t ) • Smooth-in-time ID x ( t ) . • New simple estimation method for broad class of smooth-in-time ID / hazard functions. • Biostatistics & Epidemiology methods: a little more unified?

  81. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 4. SUMMARY • Profile-specific risk (CI) functions are important. • Two paths to CI, via... • Steps-in-time S 0 ( t ) • Smooth-in-time ID x ( t ) . • New simple estimation method for broad class of smooth-in-time ID / hazard functions. • Biostatistics & Epidemiology methods: a little more unified?

  82. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 4. SUMMARY • Profile-specific risk (CI) functions are important. • Two paths to CI, via... • Steps-in-time S 0 ( t ) • Smooth-in-time ID x ( t ) . • New simple estimation method for broad class of smooth-in-time ID / hazard functions. • Biostatistics & Epidemiology methods: a little more unified?

  83. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 4. SUMMARY • Profile-specific risk (CI) functions are important. • Two paths to CI, via... • Steps-in-time S 0 ( t ) • Smooth-in-time ID x ( t ) . • New simple estimation method for broad class of smooth-in-time ID / hazard functions. • Biostatistics & Epidemiology methods: a little more unified?

  84. Introduction Smooth-in-time hazard functions How we fit fully-parametric hazard model Illustration Comments/Summary 4. SUMMARY • Profile-specific risk (CI) functions are important. • Two paths to CI, via... • Steps-in-time S 0 ( t ) • Smooth-in-time ID x ( t ) . • New simple estimation method for broad class of smooth-in-time ID / hazard functions. • Biostatistics & Epidemiology methods: a little more unified?

Recommend

Fitting smooth-in-time prognostic risk functions via logistic regression James A. Hanley 1 Olli S.

Fitting smooth-in-time prognostic risk functions via logistic regression James A. Hanley 1 Olli S.

Introduction The 2 existing approaches How we fit fully-parametric model Illustration Discussion Summary Fitting smooth-in-time prognostic risk functions via logistic regression James A. Hanley 1 Olli S. Miettinen 1 1 Department of

677 views • 40 slides
Case-base sampling for fitting and validating prognostic models Workshop on Statistical Issues in

Case-base sampling for fitting and validating prognostic models Workshop on Statistical Issues in

Case-base sampling for fitting and validating prognostic models Workshop on Statistical Issues in Biomarker and Drug Co-development Fields Institute, Toronto Olli Saarela Dalla Lana School of Public Health, University of Toronto November 8,

921 views • 55 slides
The ELN risk classification has prognostic relevance also in

The ELN risk classification has prognostic relevance also in

The ELN risk classification has prognostic relevance also in elderly patients with secondary acute myeloid leukemia and may support treatment decisions. A

389 views • 17 slides
Functions and Data Fitting COMPSCI 371D Machine Learning COMPSCI 371D Machine Learning

Functions and Data Fitting COMPSCI 371D Machine Learning COMPSCI 371D Machine Learning

Functions and Data Fitting COMPSCI 371D Machine Learning COMPSCI 371D Machine Learning Functions and Data Fitting 1 / 17 Outline 1 Functions 2 Features 3 Polynomial Fitting: Univariate Least Squares Fitting Choosing a Degree 4

397 views • 17 slides
Over fitting distribution functions over Bayesian Regression / " ' i diggllloise dist

Over fitting distribution functions over Bayesian Regression / " ' i diggllloise dist

Over fitting Under fitting and - Under fitting Over fitting distribution functions over Bayesian Regression / " ' i diggllloise dist ' f. fnpc f) fcxn ) - ,N En n=1 + En~ PIE yn = ... , of Basis Function

724 views • 30 slides
Smooth Shape-Aware Functions with Controlled Extrema Alec Jacobson 1 1 ETH Zurich Tino Weinkauf 2

Smooth Shape-Aware Functions with Controlled Extrema Alec Jacobson 1 1 ETH Zurich Tino Weinkauf 2

Smooth Shape-Aware Functions with Controlled Extrema Alec Jacobson 1 1 ETH Zurich Tino Weinkauf 2 2 MPI Saarbrcken Olga Sorkine 1 August 9, 2012 Real-time deformation relies on smooth, shape-aware functions input shape + handles August 9,

1.06k views • 87 slides
Identifying High-Risk Patients in Follicular Lymphoma with a new prognostic score Federico

Identifying High-Risk Patients in Follicular Lymphoma with a new prognostic score Federico

Identifying High-Risk Patients in Follicular Lymphoma with a new prognostic score Federico Mattiello* BBS Seminar, 1 November 2019, Basel *on behalf of the EDIS-NHL team GALLIUM clinical trial improved outcomes med.PFS 8 years with

388 views • 17 slides
Estimating Differential Quantities using Polynomial fitting of Osculating Jets

Estimating Differential Quantities using Polynomial fitting of Osculating Jets

Estimating Differential Quantities using Polynomial fitting of Osculating Jets Frederic.Cazals@inria.fr Marc.Pouget@inria.fr 1 Smooth surfaces, point clouds, meshes, Differential quantities Smooth surfaces &

521 views • 19 slides
Stochastic Optimization for DC Functions and Non-smooth Non-convex Regularizers with

Stochastic Optimization for DC Functions and Non-smooth Non-convex Regularizers with

Stochastic Optimization for DC Functions and Non-smooth Non-convex Regularizers with Non-asymptotic Convergence Yi Xu 1 , Qi Qi 1 , Qihang Lin 1 , Rong Jin 2 , Tianbao Yang 1 1. The University of Iowa 2. Damo Academy at Alibaba June 12, 2019

245 views • 11 slides
Week 2 Video 5 Cross-Validation and Over-Fitting Over-Fitting Ive mentioned over-fitting a

Week 2 Video 5 Cross-Validation and Over-Fitting Over-Fitting Ive mentioned over-fitting a

Week 2 Video 5 Cross-Validation and Over-Fitting Over-Fitting Ive mentioned over-fitting a few times during the last few weeks Fitting to the noise as well as the signal Over-Fitting 25 25 20 20 15 15 10 10 5 5 0 0 0 5 10

796 views • 25 slides
Deriving Prognostic Continuous Time Bayesian Networks from D-matrices Logan Perreault, Monica

Deriving Prognostic Continuous Time Bayesian Networks from D-matrices Logan Perreault, Monica

Deriving Prognostic Continuous Time Bayesian Networks from D-matrices Logan Perreault, Monica Thornton, Shane Strasser, and John W. Sheppard Montana State University Motivation Diagnosis and Prognosis with ATS Given: A UUT to be tested by an

559 views • 18 slides
Announcements Midterm review: next Wed Oct 4, 12-1 pm, ENS 31NQ Lecture 9: Fitting, Contours

Announcements Midterm review: next Wed Oct 4, 12-1 pm, ENS 31NQ Lecture 9: Fitting, Contours

Announcements Midterm review: next Wed Oct 4, 12-1 pm, ENS 31NQ Lecture 9: Fitting, Contours Thursday, Sept 27 Last time Today Fitting shape patterns with the Hough Fitting lines (brief) transform and generalized Hough

562 views • 16 slides
PROGNOSTIC AND PREDICTIVE BIOMARKERS IN NSCLC Federico Cappuzzo Istituto Toscano Tumori

PROGNOSTIC AND PREDICTIVE BIOMARKERS IN NSCLC Federico Cappuzzo Istituto Toscano Tumori

PROGNOSTIC AND PREDICTIVE BIOMARKERS IN NSCLC Federico Cappuzzo Istituto Toscano Tumori Ospedale Civile-Livorno Italy Prognostic versus predictive Prognostic: In presence of the biomarker patient outcome independent of the

398 views • 36 slides
Prognostic Markers can Guide Therapy? Alan K Burnett School of Medicine Cardiff University NO

Prognostic Markers can Guide Therapy? Alan K Burnett School of Medicine Cardiff University NO

Prognostic Markers can Guide Therapy? Alan K Burnett School of Medicine Cardiff University NO (t yet) Prognostic is Not Necessarily Predictive Prognostic Factor Evolution Multivariate analysis Validation on independent data Test

483 views • 36 slides
Reconstruction of Smooth 3D Color Functions from Keypoints: Application to Lossy Compression and

Reconstruction of Smooth 3D Color Functions from Keypoints: Application to Lossy Compression and

Reconstruction of Smooth 3D Color Functions from Keypoints: Application to Lossy Compression and Examplar-Based Generation of Color LUTs D. Tschumperl C. Porquet A. Mahboubi Normandie Univ, UNICAEN, ENSICAEN, CNRS GREYC, F-14000 Caen,

751 views • 32 slides
SGD without Replacement: Sharper Rates for General Smooth Convex Functions Dheeraj Nagaraj

SGD without Replacement: Sharper Rates for General Smooth Convex Functions Dheeraj Nagaraj

SGD without Replacement: Sharper Rates for General Smooth Convex Functions Dheeraj Nagaraj Massachusetts Institute of Technology June 12, 2019 Joint work with Praneeth Netrapalli and Prateek Jain (MSR India) 1 / 12 Overview Introduction 1

145 views • 12 slides
A Metric for the Prognostic Outreach of Scenarios Learning from the Past to Establish a

A Metric for the Prognostic Outreach of Scenarios Learning from the Past to Establish a

Spatio-Temporal Uncertainty Assessment of GHG Emission Inventories with the Specific Focus on Austria and Ukraine: Learning in Space and Time and into the Future A Metric for the Prognostic Outreach of Scenarios Learning from the Past to

514 views • 34 slides
Last time What are grouping problems in vision? Inspiration from human perception

Last time What are grouping problems in vision? Inspiration from human perception

2/13/2017 Fitting : Voting and the Hough Transform Tues Feb 14 Kristen Grauman UT Austin Today Grouping : wrap up clustering algorithms See slides from last time Fitting : introduction to voting Slide credit: Kristen Grauman 1

479 views • 12 slides
Recall: integrating functions over curves Let C be a smooth curve in R 3 parametrized by r ( t ), a

Recall: integrating functions over curves Let C be a smooth curve in R 3 parametrized by r ( t ), a

Recall: integrating functions over curves Let C be a smooth curve in R 3 parametrized by r ( t ), a t b . Then b | r ( t ) | dt . Length( C ) = a Furthermore, if g is a continuous function on C , then b g ( r ( t

330 views • 7 slides
Lecture 11 Fitting ARIMA Models 10/10/2018 1 Model Fitting Fitting ARIMA For an

Lecture 11 Fitting ARIMA Models 10/10/2018 1 Model Fitting Fitting ARIMA For an

Lecture 11 Fitting ARIMA Models 10/10/2018 1 Model Fitting Fitting ARIMA For an (, , ) model Requires that the data be stationary after differencing Handling is straight forward, just difference the

682 views • 46 slides
Fitting Nonlinear Models to Data SI Model The SI model we discussed before is often written dS

Fitting Nonlinear Models to Data SI Model The SI model we discussed before is often written dS

Fitting Nonlinear Models to Data SI Model The SI model we discussed before is often written dS / dt = pS I d I / dt = pS I where S is the susceptible population those at risk to become infected at a given time and I is the

274 views • 9 slides
Quantization causes waves Smooth finitely computable functions are affine Vladimir Anashin

Quantization causes waves Smooth finitely computable functions are affine Vladimir Anashin

Quantization causes waves Smooth finitely computable functions are affine Vladimir Anashin Faculty of Computational Mathematics and Cybernetics Faculty of Physics Lomonosov Moscow State University ***************************** Institute of

1.47k views • 101 slides
Lecture 9: Fitting, Contours Thursday, Sept 27 Announcements Midterm review: next Wed Oct

Lecture 9: Fitting, Contours Thursday, Sept 27 Announcements Midterm review: next Wed Oct

Lecture 9: Fitting, Contours Thursday, Sept 27 Announcements Midterm review: next Wed Oct 4, 12-1 pm, ENS 31NQ Last time Fitting shape patterns with the Hough transform and generalized Hough transform Today Fitting lines

1.13k views • 94 slides
T -maturity ZCB; time- t price denoted P ( t ; T ) . As a fct of T : Smooth. As a fct of t :

T -maturity ZCB; time- t price denoted P ( t ; T ) . As a fct of T : Smooth. As a fct of t :

T -maturity ZCB; time- t price denoted P ( t ; T ) . As a fct of T : Smooth. As a fct of t : Erratic (Ito-process). Continuously compounded ZC yield y ( t, ) is defined by P ( t ; t + ) = exp( y ( t ; )) y ( t ; ) = ln P (

269 views • 16 slides

More recommend