copulas for neural and behavioral parallel systems hans
play

Copulas for neural and behavioral parallel systems Hans Colonius - PowerPoint PPT Presentation

Copulas for neural and behavioral parallel systems Hans Colonius Department f ur Psychologie Universit at Oldenburg Purdue Winer Memorial Lectures Probability and Contextuality November 2018 1 / 75 Outline Preliminaries Coupling:


  1. Defining the copula Definition An n - copula is an n -variate distribution function with univariate margins uniformly distributed on [0 , 1]. ◮ There are many equivalent definitions of copula. ◮ One of them, in the case n = 2, is the following: Definition A function C : [0 , 1] × [0 , 1] → [0 , 1] is a 2 − copula if, and only if, it satisfies 1. C (0 , t ) = C ( t , 0) = 0, for every t ∈ [0 , 1] (groundedness); 2. C (1 , t ) = C ( t , 1) = t , for every t ∈ [0 , 1] (uniform marginals); 3. for all a 1 , a 2 , b 1 , b 2 ∈ [0 , 1], with a 1 ≤ b 1 and a 2 ≤ b 2 , ( ∗ ) C ( a 1 , a 2 ) − C ( a 1 , b 2 ) − C ( b 1 , a 2 ) + C ( b 1 , b 2 ) ≥ 0 . ( ∗ ) is called 2-increasing, supermodularity . 12 / 75

  2. Defining the copula Definition An n - copula is an n -variate distribution function with univariate margins uniformly distributed on [0 , 1]. ◮ There are many equivalent definitions of copula. ◮ One of them, in the case n = 2, is the following: Definition A function C : [0 , 1] × [0 , 1] → [0 , 1] is a 2 − copula if, and only if, it satisfies 1. C (0 , t ) = C ( t , 0) = 0, for every t ∈ [0 , 1] (groundedness); 2. C (1 , t ) = C ( t , 1) = t , for every t ∈ [0 , 1] (uniform marginals); 3. for all a 1 , a 2 , b 1 , b 2 ∈ [0 , 1], with a 1 ≤ b 1 and a 2 ≤ b 2 , ( ∗ ) C ( a 1 , a 2 ) − C ( a 1 , b 2 ) − C ( b 1 , a 2 ) + C ( b 1 , b 2 ) ≥ 0 . ( ∗ ) is called 2-increasing, supermodularity . 12 / 75

  3. Defining the copula Definition An n - copula is an n -variate distribution function with univariate margins uniformly distributed on [0 , 1]. ◮ There are many equivalent definitions of copula. ◮ One of them, in the case n = 2, is the following: Definition A function C : [0 , 1] × [0 , 1] → [0 , 1] is a 2 − copula if, and only if, it satisfies 1. C (0 , t ) = C ( t , 0) = 0, for every t ∈ [0 , 1] (groundedness); 2. C (1 , t ) = C ( t , 1) = t , for every t ∈ [0 , 1] (uniform marginals); 3. for all a 1 , a 2 , b 1 , b 2 ∈ [0 , 1], with a 1 ≤ b 1 and a 2 ≤ b 2 , ( ∗ ) C ( a 1 , a 2 ) − C ( a 1 , b 2 ) − C ( b 1 , a 2 ) + C ( b 1 , b 2 ) ≥ 0 . ( ∗ ) is called 2-increasing, supermodularity . 12 / 75

  4. Sklar’s Theorem Theorem (Sklar’s Theorem, 1959) Let F ( x 1 , . . . , x n ) be an n-variate distribution function with margins F 1 ( x 1 ) , . . . , F n ( x n ) ; then there exists an n-copula C : [0 , 1] n − → [0 , 1] that satisfies ( x 1 , . . . x n ) ∈ ℜ n . F ( x 1 , . . . , x n ) = C ( F 1 ( x 1 ) , . . . , F n ( x n )) , If all univariate margins F 1 , . . . , F n are continuous, then the copula is unique. If F − 1 1 , . . . , F − 1 are the quantile functions of the margins, n then for any ( u 1 , . . . , u n ) ∈ [0 , 1] n C ( u 1 , . . . , u n ) = F ( F − 1 1 ( u 1 ) , . . . , F − 1 n ( u n )) . Note: The copula can be considered ‘independent‘ of the univariate margins. 13 / 75

  5. Sklar’s Theorem Theorem (Sklar’s Theorem, 1959) Let F ( x 1 , . . . , x n ) be an n-variate distribution function with margins F 1 ( x 1 ) , . . . , F n ( x n ) ; then there exists an n-copula C : [0 , 1] n − → [0 , 1] that satisfies ( x 1 , . . . x n ) ∈ ℜ n . F ( x 1 , . . . , x n ) = C ( F 1 ( x 1 ) , . . . , F n ( x n )) , If all univariate margins F 1 , . . . , F n are continuous, then the copula is unique. If F − 1 1 , . . . , F − 1 are the quantile functions of the margins, n then for any ( u 1 , . . . , u n ) ∈ [0 , 1] n C ( u 1 , . . . , u n ) = F ( F − 1 1 ( u 1 ) , . . . , F − 1 n ( u n )) . Note: The copula can be considered ‘independent‘ of the univariate margins. 13 / 75

  6. How to prove that a function is a copula ◮ Find a suitable probabilistic model (i.e., random vector) whose distribution function is concentrated on [0 , 1] n and has uniform marginals; or, ◮ prove that properties (1) groundedness, (2) uniform marginals, and (3) n -increasing (supermodularity) are satisfied. 14 / 75

  7. How to prove that a function is a copula ◮ Find a suitable probabilistic model (i.e., random vector) whose distribution function is concentrated on [0 , 1] n and has uniform marginals; or, ◮ prove that properties (1) groundedness, (2) uniform marginals, and (3) n -increasing (supermodularity) are satisfied. 14 / 75

  8. Properties of copulas Sklar’s theorem shows that copulas remain invariant under strictly increasing transformations of the underlying random variables. It is possible to construct a wide range of multivariate distributions by choosing the marginal distributions and a suitable copula. 15 / 75

  9. Example 1: Gumbel’s bivariate exponential copula Let H θ be the joint distribution function given by 1 − e − x − e − y + e − ( x + y + θ xy ) � if x ≥ 0 , y ≥ 0; H θ ( x , y ) = 0 , otherwise; where θ is a parameter in [0 , 1]. Then the marginals are exponentials, with quantile functions F − 1 ( u ) = − ln(1 − u ) and G − 1 ( v ) = − ln(1 − v ) for u , v ∈ [0 , 1]. The corresponding copula is C θ ( u , v ) = u + v − 1 + (1 − u )(1 − v ) e − θ ln(1 − u ) ln(1 − v ) . 16 / 75

  10. Example 1: Gumbel’s bivariate exponential copula Let H θ be the joint distribution function given by 1 − e − x − e − y + e − ( x + y + θ xy ) � if x ≥ 0 , y ≥ 0; H θ ( x , y ) = 0 , otherwise; where θ is a parameter in [0 , 1]. Then the marginals are exponentials, with quantile functions F − 1 ( u ) = − ln(1 − u ) and G − 1 ( v ) = − ln(1 − v ) for u , v ∈ [0 , 1]. The corresponding copula is C θ ( u , v ) = u + v − 1 + (1 − u )(1 − v ) e − θ ln(1 − u ) ln(1 − v ) . 16 / 75

  11. Example 1: Gumbel’s bivariate exponential copula Let H θ be the joint distribution function given by 1 − e − x − e − y + e − ( x + y + θ xy ) � if x ≥ 0 , y ≥ 0; H θ ( x , y ) = 0 , otherwise; where θ is a parameter in [0 , 1]. Then the marginals are exponentials, with quantile functions F − 1 ( u ) = − ln(1 − u ) and G − 1 ( v ) = − ln(1 − v ) for u , v ∈ [0 , 1]. The corresponding copula is C θ ( u , v ) = u + v − 1 + (1 − u )(1 − v ) e − θ ln(1 − u ) ln(1 − v ) . 16 / 75

  12. Example 2: Bivariate extreme value distribution Let X and Y be random variables with a joint distribution given by H θ ( x , y ) = exp[ − ( e − θ x + e − θ y ) 1 /θ ] for all x , y ∈ ¯ ℜ , where θ ≥ 1. The corresponding Gumbel-Hougaard copula is given by � ( − ln u ) θ + ( − ln v ) θ � 1 /θ � � C θ ( u , v ) = exp − . 17 / 75

  13. Example 2: Bivariate extreme value distribution Let X and Y be random variables with a joint distribution given by H θ ( x , y ) = exp[ − ( e − θ x + e − θ y ) 1 /θ ] for all x , y ∈ ¯ ℜ , where θ ≥ 1. The corresponding Gumbel-Hougaard copula is given by � ( − ln u ) θ + ( − ln v ) θ � 1 /θ � � C θ ( u , v ) = exp − . 17 / 75

  14. Plotting Gumbel-Hougaard copula with different marginals UniformDistribution �� 0, 1 �� 1.0 2 0.5 1 0.0 0.0 0 � 1 0.5 � 2 1.0 ExponentialDistribution � 2 � 1.0 2 0.5 1 0.0 18 / 75 0.0 0 � 1 0.5 � 2 1.0

  15. Plotting Gumbel-Hougaard copula with different marginals NormalDistribution � 0, 1 � 1.0 2 0.5 1 0.0 0.0 0 � 1 0.5 � 2 1.0 LaplaceDistribution � 0, 1 � 0.8 2 0.6 0.4 1 0.2 0.0 19 / 75 0.0 0 � 1 0.5 � 2 1.0

  16. Plotting Gumbel-Hougaard copula with different marginals GumbelDistribution � 1, 2 � 0.8 0.6 2 0.4 1 0.2 0.0 0.0 0 � 1 0.5 � 2 1.0 WeibullDistribution � 2, 1 � 1.0 2 0.5 1 0.0 20 / 75 0.0 0 � 1 0.5 � 2 1.0

  17. Fr´ echet-Hoeffding bounds and copulas These will be central topics in the applications presented here. 21 / 75

  18. Fr´ echet-Hoeffding copulas Let C ( u , v ) be a 2-copula; then, for u , v ∈ [0 , 1], W ( u , v ) ≡ max { u + v − 1 , 0 } ≤ C ( u , v ) ≤ min { u , v } ≡ M ( u , v ) , and M and W are also copulas, the upper and lower Fr´ echet-Hoeffding copula . The dependence properties translate into: For W , P ( U + V = 1) = 1, and for M , P ( U = V ) = 1. 22 / 75

  19. Fr´ echet-Hoeffding copulas Let C ( u , v ) be a 2-copula; then, for u , v ∈ [0 , 1], W ( u , v ) ≡ max { u + v − 1 , 0 } ≤ C ( u , v ) ≤ min { u , v } ≡ M ( u , v ) , and M and W are also copulas, the upper and lower Fr´ echet-Hoeffding copula . The dependence properties translate into: For W , P ( U + V = 1) = 1, and for M , P ( U = V ) = 1. 22 / 75

  20. Copulas and order statistics: diagonal section Maximum Let U 1 , U 2 , . . . , U n be random variables defined on the same prob. space., having uniform distribution on (0 , 1) and C as their distribution function; then, for every t ∈ [0 , 1]   n � P (max { U 1 , U 2 , . . . , U n } ) ≤ t ) = P { U j ≤ t }   j =1 = C ( t , t , . . . , t ) δ C ( t ) := C ( t , t , . . . , t ) is called the diagonal section of the copula C . 23 / 75

  21. Copulas and order statistics: diagonal section Maximum Let U 1 , U 2 , . . . , U n be random variables defined on the same prob. space., having uniform distribution on (0 , 1) and C as their distribution function; then, for every t ∈ [0 , 1]   n � P (max { U 1 , U 2 , . . . , U n } ) ≤ t ) = P { U j ≤ t }   j =1 = C ( t , t , . . . , t ) δ C ( t ) := C ( t , t , . . . , t ) is called the diagonal section of the copula C . 23 / 75

  22. Copulas and order statistics: diagonal section Minimum For n = 2, it follows P (min { U , V } ) ≤ t ) = P ( U ≤ t ) + P ( V ≤ t ) − P ( U ≤ t ∩ V ≤ t ) = 2 t − δ C ( t ) . Thus, determining a bivariate copula C with prescribed diagonal section δ is equivalent to determining a random vector ( U , V ) such that ( U , V ) ∼ C and the marginal distribution functions of the order statistics of ( U , V ) are known. Note: the diagonal of a copula does not uniquely determine the underlying copula. 24 / 75

  23. Copulas and order statistics: diagonal section Minimum For n = 2, it follows P (min { U , V } ) ≤ t ) = P ( U ≤ t ) + P ( V ≤ t ) − P ( U ≤ t ∩ V ≤ t ) = 2 t − δ C ( t ) . Thus, determining a bivariate copula C with prescribed diagonal section δ is equivalent to determining a random vector ( U , V ) such that ( U , V ) ∼ C and the marginal distribution functions of the order statistics of ( U , V ) are known. Note: the diagonal of a copula does not uniquely determine the underlying copula. 24 / 75

  24. Copulas and order statistics: diagonal section Minimum For n = 2, it follows P (min { U , V } ) ≤ t ) = P ( U ≤ t ) + P ( V ≤ t ) − P ( U ≤ t ∩ V ≤ t ) = 2 t − δ C ( t ) . Thus, determining a bivariate copula C with prescribed diagonal section δ is equivalent to determining a random vector ( U , V ) such that ( U , V ) ∼ C and the marginal distribution functions of the order statistics of ( U , V ) are known. Note: the diagonal of a copula does not uniquely determine the underlying copula. 24 / 75

  25. Application to parallel systems We loosely define Parallel system: “Existence of one or more multivariate distributions on possibly different probability spaces, where random variables relate to certain behavioral or neural processes”. 25 / 75

  26. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 26 / 75

  27. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 26 / 75

  28. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 26 / 75

  29. The multisensory paradigm (behavioral version) ◮ Unimodal condition: a stimulus of a single modality (visual, auditory, tactile) is presented and the participant is asked to respond (by button press or eye movement) as quickly as possible upon detecting the stimulus ( reaction time, RT, task ). ◮ Bi- or trimodal condition: stimuli from two or three modalities are presented (nearly) simultaneously and the participant is asked to respond as quickly as possible upon detecting a stimulus of any modality ( redundant signals task ) 27 / 75

  30. The multisensory paradigm (behavioral version) ◮ We refer to V , A , T as the unimodal context where visual, auditory, or tactile stimuli are presented, resp. Simlarly, VA denotes a bimodal (visual-auditory) context, etc. ◮ For each stimulus, or stimulus combination, we observe samples from a random variable representing the reaction time measured in any given trial. Let F V ( t ) , F A ( t ) , F VA ( t ) , . . . denote the distribution functions of reaction time in a unimodal visual, auditory, or a visual-auditory context, etc. when a specific stimulus (combination) is presented. 28 / 75

  31. The multisensory paradigm (behavioral version) ◮ We refer to V , A , T as the unimodal context where visual, auditory, or tactile stimuli are presented, resp. Simlarly, VA denotes a bimodal (visual-auditory) context, etc. ◮ For each stimulus, or stimulus combination, we observe samples from a random variable representing the reaction time measured in any given trial. Let F V ( t ) , F A ( t ) , F VA ( t ) , . . . denote the distribution functions of reaction time in a unimodal visual, auditory, or a visual-auditory context, etc. when a specific stimulus (combination) is presented. 28 / 75

  32. Existence of couplings ◮ Each context V , A , VA refers to a different event space ( σ -algebra), so, from an empirical point of view, no couplings between the reaction time random variables in these different conditions necessarily exist. ◮ A common assumption, often not stated explicitly, is that there exists a coupling between visual and auditory RT, for example, such that the margins of the bivariate distribution H VA are equal to F V and F A . ◮ Given a coupling exists, an assumption on how H VA is related to the margins of F V and F A , is required ( ⇒ copula ?). ◮ In principle, this could be tested empirically. However, the margins of H VA are not observable, only the distribution function of RTs in the bimodal context, F VA , is. 29 / 75

  33. Existence of couplings ◮ Each context V , A , VA refers to a different event space ( σ -algebra), so, from an empirical point of view, no couplings between the reaction time random variables in these different conditions necessarily exist. ◮ A common assumption, often not stated explicitly, is that there exists a coupling between visual and auditory RT, for example, such that the margins of the bivariate distribution H VA are equal to F V and F A . ◮ Given a coupling exists, an assumption on how H VA is related to the margins of F V and F A , is required ( ⇒ copula ?). ◮ In principle, this could be tested empirically. However, the margins of H VA are not observable, only the distribution function of RTs in the bimodal context, F VA , is. 29 / 75

  34. Existence of couplings ◮ Each context V , A , VA refers to a different event space ( σ -algebra), so, from an empirical point of view, no couplings between the reaction time random variables in these different conditions necessarily exist. ◮ A common assumption, often not stated explicitly, is that there exists a coupling between visual and auditory RT, for example, such that the margins of the bivariate distribution H VA are equal to F V and F A . ◮ Given a coupling exists, an assumption on how H VA is related to the margins of F V and F A , is required ( ⇒ copula ?). ◮ In principle, this could be tested empirically. However, the margins of H VA are not observable, only the distribution function of RTs in the bimodal context, F VA , is. 29 / 75

  35. Existence of couplings ◮ Each context V , A , VA refers to a different event space ( σ -algebra), so, from an empirical point of view, no couplings between the reaction time random variables in these different conditions necessarily exist. ◮ A common assumption, often not stated explicitly, is that there exists a coupling between visual and auditory RT, for example, such that the margins of the bivariate distribution H VA are equal to F V and F A . ◮ Given a coupling exists, an assumption on how H VA is related to the margins of F V and F A , is required ( ⇒ copula ?). ◮ In principle, this could be tested empirically. However, the margins of H VA are not observable, only the distribution function of RTs in the bimodal context, F VA , is. 29 / 75

  36. The (possibly non-independent) race model The model studied most often is the (possibly non-independent) race model : Let V and A be the random reaction times in unimodal conditions V and A , with distribution functions F V ( t ) and F A ( t ), resp. ◮ Assume bimodal RT is determined by the “winner” of the race between the modalities: F VA ( t ) = P ( V ≤ t or A ≤ t ) ◮ Then F VA ( t ) = F V ( t ) + F A ( t ) − H VA ( t , t ) , with H VA ( t , t ) = P ( V ≤ t and A ≤ t ). 30 / 75

  37. The (possibly non-independent) race model The model studied most often is the (possibly non-independent) race model : Let V and A be the random reaction times in unimodal conditions V and A , with distribution functions F V ( t ) and F A ( t ), resp. ◮ Assume bimodal RT is determined by the “winner” of the race between the modalities: F VA ( t ) = P ( V ≤ t or A ≤ t ) ◮ Then F VA ( t ) = F V ( t ) + F A ( t ) − H VA ( t , t ) , with H VA ( t , t ) = P ( V ≤ t and A ≤ t ). 30 / 75

  38. The (possibly non-independent) race model The model studied most often is the (possibly non-independent) race model : Let V and A be the random reaction times in unimodal conditions V and A , with distribution functions F V ( t ) and F A ( t ), resp. ◮ Assume bimodal RT is determined by the “winner” of the race between the modalities: F VA ( t ) = P ( V ≤ t or A ≤ t ) ◮ Then F VA ( t ) = F V ( t ) + F A ( t ) − H VA ( t , t ) , with H VA ( t , t ) = P ( V ≤ t and A ≤ t ). 30 / 75

  39. The (possibly non-independent) race model ◮ from the Fr´ echet bounds, max { F V ( t ) + F A ( t ) − 1 , 0 } ≤ H VA ( t , t ) ≤ min { F V ( t ) , F A ( t ) } . ◮ Inserting and rearranging yields max { F V ( t ) , F A ( t ) } ≤ F VA ( t ) ≤ min { F V ( t )+ F A ( t ) , 1 } , t ≥ 0 , (‘race model inequality’, Miller 1982) ◮ The upper bound corresponds to maximal negative dependence between V and A , the lower bound to maximal positive dependence. ◮ Empirical violation of the upper bound (occurring only for small enough t ) is interpreted as evidence against the race mechanism (“bimodal RT faster than predictable from unimodal conditions”). 31 / 75

  40. The (possibly non-independent) race model ◮ from the Fr´ echet bounds, max { F V ( t ) + F A ( t ) − 1 , 0 } ≤ H VA ( t , t ) ≤ min { F V ( t ) , F A ( t ) } . ◮ Inserting and rearranging yields max { F V ( t ) , F A ( t ) } ≤ F VA ( t ) ≤ min { F V ( t )+ F A ( t ) , 1 } , t ≥ 0 , (‘race model inequality’, Miller 1982) ◮ The upper bound corresponds to maximal negative dependence between V and A , the lower bound to maximal positive dependence. ◮ Empirical violation of the upper bound (occurring only for small enough t ) is interpreted as evidence against the race mechanism (“bimodal RT faster than predictable from unimodal conditions”). 31 / 75

  41. The (possibly non-independent) race model ◮ from the Fr´ echet bounds, max { F V ( t ) + F A ( t ) − 1 , 0 } ≤ H VA ( t , t ) ≤ min { F V ( t ) , F A ( t ) } . ◮ Inserting and rearranging yields max { F V ( t ) , F A ( t ) } ≤ F VA ( t ) ≤ min { F V ( t )+ F A ( t ) , 1 } , t ≥ 0 , (‘race model inequality’, Miller 1982) ◮ The upper bound corresponds to maximal negative dependence between V and A , the lower bound to maximal positive dependence. ◮ Empirical violation of the upper bound (occurring only for small enough t ) is interpreted as evidence against the race mechanism (“bimodal RT faster than predictable from unimodal conditions”). 31 / 75

  42. The (possibly non-independent) race model ◮ from the Fr´ echet bounds, max { F V ( t ) + F A ( t ) − 1 , 0 } ≤ H VA ( t , t ) ≤ min { F V ( t ) , F A ( t ) } . ◮ Inserting and rearranging yields max { F V ( t ) , F A ( t ) } ≤ F VA ( t ) ≤ min { F V ( t )+ F A ( t ) , 1 } , t ≥ 0 , (‘race model inequality’, Miller 1982) ◮ The upper bound corresponds to maximal negative dependence between V and A , the lower bound to maximal positive dependence. ◮ Empirical violation of the upper bound (occurring only for small enough t ) is interpreted as evidence against the race mechanism (“bimodal RT faster than predictable from unimodal conditions”). 31 / 75

  43. Race Model Inequality Test The grey area between the upper bound F A + F V and the bimodal RT distribution F VA is taken as a measure of the amount of violation of the race model inequality. This area is the (sample estimate of the) expected value of min( A , V ): E ( − ) [min( V , A )] , under maximal negative dependence between V and A (Colonius & Diederich PsyRev Gondan & Minakata 2016 2006). Estimation is straightforward. 32 / 75

  44. Crossmodal response enhancement (CRE) ◮ We will use maximal negative probability summation to define a new measure of crossmodal response enhancemen for RT. ◮ Response enhancement in RT means “faster average responses”: CRE RT = min { E [ RT V ] , E [ RT A ] } − E [ RT VA ] × 100 min { E [ RT V ] , E [ RT A ] 33 / 75

  45. Crossmodal response enhancement (CRE) ◮ We will use maximal negative probability summation to define a new measure of crossmodal response enhancemen for RT. ◮ Response enhancement in RT means “faster average responses”: CRE RT = min { E [ RT V ] , E [ RT A ] } − E [ RT VA ] × 100 min { E [ RT V ] , E [ RT A ] 33 / 75

  46. The new measure of CRE in RT 1. Replace the traditional CRE RT = min { E [ RT V ] , E [ RT A ] } − E [ RT VA ] × 100 min { E [ RT V ] , E [ RT A ] by CRE RT ( − ) = E ( − ) [min( V , A )] − E [ RT VA ] × 100 E ( − ) [min( V , A )] 2. CRE RT ( − ) ≤ CRE RT 34 / 75

  47. The multisensory paradigm (neural version) 35 / 75

  48. CRE at the level of a single neuron Response strength: the absolute number of impulses (spikes) registered within in a fixed time interval after stimulus presentation Stein et al., Nat Rev Neurosci 2014 CRE = CM VA − max { UM V , UM A } × 100 , max { UM V , UM A } CRE = 123[%] 36 / 75

  49. CRE at the level of a single neuron Response strength: the absolute number of impulses (spikes) registered within in a fixed time interval after stimulus presentation Stein et al., Nat Rev Neurosci 2014 CRE = CM VA − max { UM V , UM A } × 100 , max { UM V , UM A } CRE = 123[%] 36 / 75

  50. CRE at the level of a single neuron Response strength: the absolute number of impulses (spikes) registered within in a fixed time interval after stimulus presentation Stein et al., Nat Rev Neurosci 2014 CRE = CM VA − max { UM V , UM A } × 100 , max { UM V , UM A } CRE = 123[%] 36 / 75

  51. CRE at the level of a single neuron Response strength: the absolute number of impulses (spikes) registered within in a fixed time interval after stimulus presentation Stein et al., Nat Rev Neurosci 2014 CRE = CM VA − max { UM V , UM A } × 100 , max { UM V , UM A } CRE = 123[%] 36 / 75

  52. CRE at the level of a single neuron ◮ this measure is a very useful tool, but it is purely descriptive ◮ no theoretical foundation in terms of the possible operations SC neurons may perform = ⇒ Being responsive to multiple sensory modalities does not guarantee that a neuron has actually engaged in integrating its multiple sensory inputs rather than simply responding to the most salient stimulus. Multisensory computations performed by SC neurons still not fully understood 37 / 75

  53. CRE at the level of a single neuron ◮ this measure is a very useful tool, but it is purely descriptive ◮ no theoretical foundation in terms of the possible operations SC neurons may perform = ⇒ Being responsive to multiple sensory modalities does not guarantee that a neuron has actually engaged in integrating its multiple sensory inputs rather than simply responding to the most salient stimulus. Multisensory computations performed by SC neurons still not fully understood 37 / 75

  54. CRE at the level of a single neuron ◮ this measure is a very useful tool, but it is purely descriptive ◮ no theoretical foundation in terms of the possible operations SC neurons may perform = ⇒ Being responsive to multiple sensory modalities does not guarantee that a neuron has actually engaged in integrating its multiple sensory inputs rather than simply responding to the most salient stimulus. Multisensory computations performed by SC neurons still not fully understood 37 / 75

  55. CRE at the level of a single neuron ◮ this measure is a very useful tool, but it is purely descriptive ◮ no theoretical foundation in terms of the possible operations SC neurons may perform = ⇒ Being responsive to multiple sensory modalities does not guarantee that a neuron has actually engaged in integrating its multiple sensory inputs rather than simply responding to the most salient stimulus. Multisensory computations performed by SC neurons still not fully understood 37 / 75

  56. CRE for single neuron data ◮ Task: Develop a measure of crossmodal enhancement for single neuron data, in analogy to CRE RT ( − ) . ◮ N V , N A , and N VA denote random number of impulses emitted by a neuron, following unisensory (visual, auditory) and cross-modal stimulation. The traditional index is: CRE MAX = E [ N VA ] − max { E [ N V ] , E [ N A ] } × 100 . max { E [ N V ] , E [ N A ] } 38 / 75

  57. Towards a new index of CRE in neural data G V ( m ) = P V ( N V > m ) and G A ( m ) = P A ( N A > m ), m = 0 , 1 , . . . the (survivor) distribution functions of N V and N A . In analogy to the race model inequality, from the Fr´ echet-Hoeffding bounds , min { G V ( m ) , G A ( m ) } ≤ P (max { N V , N A } > m ) ≤ max { 0 , G V ( m ) + G A ( m ) − 1 } for m = 0 , 1 , . . . Again, upper and lower bounds are distribution functions (survival fcts) for random variable max { N V , N A } ! Summing over m and applying Jensen’ s inequality , we obtain 39 / 75

  58. Proposition on Coupling N V and N A Proposition For any coupling of the univariate response random variables N V and N A with expected value E [max { N V , N A } ] , max { E [ N V ] , E [ N A ] } ≤ E [max { N V , N A } ] ≤ E ( − ) [max { N V , N A } ] , where E ( − ) [max { N V , N A } ] is the expected value under maximal negative dependence between N V and N A . 40 / 75

  59. A a new index of CRE in neural data Replace the traditional index CRE MAX = E [ N VA ] − max { E [ N V ] , E [ N A ] } × 100 . max { E [ N V ] , E [ N A ] } by MAX = E [ N VA ] − E ( − ) [max { N V , N A } ] CRE ( − ) × 100 . E ( − ) [max { N V , N A } ] CRE ( − ) MAX compares the observed bimodal response E N VA with the largest bimodal response achievable by coupling the unisensory responses via negative stochastic dependence. 41 / 75

  60. An Important Consequence ◮ CRE ( − ) MAX ≤ CRE MAX ⇒ Thus, a neuron labeled as “multisensory” under CRE MAX may lose that property under CRE ( − ) MAX . 42 / 75

  61. Illustrative Example: Single SC Neuron Data * Data set from Mark Wallace (Vanderbilt U.): 84 sessions, with 15 trials each, from 3 superior colliculus neurons (cat) (Results in Colonius & Diederich, Scientific Reports , 2017.) 43 / 75

  62. Intermediate summary ◮ We suggest to replace the traditional multisensory index CRE MAX by the new one, CRE ( − ) MAX . ◮ The new one has a theoretical foundation: it measures the degree by which a neuron’s observed multisensory response surpasses the level obtainable by optimally combining the unisensory responses (assuming that the neuron simply reacts to the more salient modality in any given cross-modal trial). ◮ CRE ( − ) MAX is easy to compute and does not require any specific assumption about the distribution of the spikes. ◮ It is in straightforward analogy to a measure well-established in the domain of reaction times (“race model inequality”). ◮ It is sensitive to the variability of the data set. 44 / 75

  63. Intermediate summary ◮ We suggest to replace the traditional multisensory index CRE MAX by the new one, CRE ( − ) MAX . ◮ The new one has a theoretical foundation: it measures the degree by which a neuron’s observed multisensory response surpasses the level obtainable by optimally combining the unisensory responses (assuming that the neuron simply reacts to the more salient modality in any given cross-modal trial). ◮ CRE ( − ) MAX is easy to compute and does not require any specific assumption about the distribution of the spikes. ◮ It is in straightforward analogy to a measure well-established in the domain of reaction times (“race model inequality”). ◮ It is sensitive to the variability of the data set. 44 / 75

  64. Intermediate summary ◮ We suggest to replace the traditional multisensory index CRE MAX by the new one, CRE ( − ) MAX . ◮ The new one has a theoretical foundation: it measures the degree by which a neuron’s observed multisensory response surpasses the level obtainable by optimally combining the unisensory responses (assuming that the neuron simply reacts to the more salient modality in any given cross-modal trial). ◮ CRE ( − ) MAX is easy to compute and does not require any specific assumption about the distribution of the spikes. ◮ It is in straightforward analogy to a measure well-established in the domain of reaction times (“race model inequality”). ◮ It is sensitive to the variability of the data set. 44 / 75

  65. Intermediate summary ◮ We suggest to replace the traditional multisensory index CRE MAX by the new one, CRE ( − ) MAX . ◮ The new one has a theoretical foundation: it measures the degree by which a neuron’s observed multisensory response surpasses the level obtainable by optimally combining the unisensory responses (assuming that the neuron simply reacts to the more salient modality in any given cross-modal trial). ◮ CRE ( − ) MAX is easy to compute and does not require any specific assumption about the distribution of the spikes. ◮ It is in straightforward analogy to a measure well-established in the domain of reaction times (“race model inequality”). ◮ It is sensitive to the variability of the data set. 44 / 75

  66. Intermediate summary ◮ We suggest to replace the traditional multisensory index CRE MAX by the new one, CRE ( − ) MAX . ◮ The new one has a theoretical foundation: it measures the degree by which a neuron’s observed multisensory response surpasses the level obtainable by optimally combining the unisensory responses (assuming that the neuron simply reacts to the more salient modality in any given cross-modal trial). ◮ CRE ( − ) MAX is easy to compute and does not require any specific assumption about the distribution of the spikes. ◮ It is in straightforward analogy to a measure well-established in the domain of reaction times (“race model inequality”). ◮ It is sensitive to the variability of the data set. 44 / 75

  67. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 45 / 75

  68. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 45 / 75

  69. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 45 / 75

  70. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 45 / 75

  71. Copulas for neural and behavioral parallel systems (1) Multisensory integration: reaction times (2) Multisensory integration: spike numbers (impulse numbers) (3) Response inhibition: stop signal paradigm In (1) and (2), we present a new measure of multisensory integration that unifies behavioral and neural data. In (3), we modify a classic model of inhibitory behavior in a way to solve a paradox between behavioral and neural data. 45 / 75

  72. Response inhibition: stop signal paradigm 46 / 75

  73. Stop signal paradigm - Subjects are instructed to make a response as quickly as possible to a go signal (no-stop-signal trial) - On a minority of trials, a stop signal is presented and subjects have to inhibit the previously planned response (stop-signal trial) 47 / 75

  74. Stop signal paradigm: inhibition functions - Inhibition functions of three subjects (Logan & Cowan, 1984) - The inhibition function is determined by stop-signal delay, but it also depends strongly on RT in the go task; the probability of responding given a stop signal is lower the longer the go RT 48 / 75

  75. RT distributions with and without stop signal - Observed response time distributions for no-stop -signal trials and signal-respond trials with stop-signal delays (SSDs) of 153, 241, and 329 ms (from Logan et al 2014): - RT signal-respond < RT go - faster for shorter stop-signal delays than for longer ones. 49 / 75

  76. RT distributions with and without stop signal - Observed response time distributions for no-stop -signal trials and signal-respond trials with stop-signal delays (SSDs) of 153, 241, and 329 ms (from Logan et al 2014): - RT signal-respond < RT go - faster for shorter stop-signal delays than for longer ones. 49 / 75

  77. The general race model 50 / 75

  78. The general race model (1) ◮ Distinguish two different experimental conditions termed context GO , where only a go signal is presented, and context STOP , where a stop signal is presented in addition. ◮ In STOP , let T go and T stop denote the random processing time for the go and the stop signal, respectively, with (unobservable !) bivariate distribution function H ( s , t ) = Pr( T go ≤ s , T stop ≤ t ) , for all s , t ≥ 0. ◮ The marginal distributions of H ( s , t ) are denoted as F go ( s ) = Pr( T go ≤ s , T stop < ∞ ) F stop ( t ) = Pr( T go < ∞ , T stop ≤ t ) . 51 / 75

  79. The general race model (1) ◮ Distinguish two different experimental conditions termed context GO , where only a go signal is presented, and context STOP , where a stop signal is presented in addition. ◮ In STOP , let T go and T stop denote the random processing time for the go and the stop signal, respectively, with (unobservable !) bivariate distribution function H ( s , t ) = Pr( T go ≤ s , T stop ≤ t ) , for all s , t ≥ 0. ◮ The marginal distributions of H ( s , t ) are denoted as F go ( s ) = Pr( T go ≤ s , T stop < ∞ ) F stop ( t ) = Pr( T go < ∞ , T stop ≤ t ) . 51 / 75

  80. The general race model (1) ◮ Distinguish two different experimental conditions termed context GO , where only a go signal is presented, and context STOP , where a stop signal is presented in addition. ◮ In STOP , let T go and T stop denote the random processing time for the go and the stop signal, respectively, with (unobservable !) bivariate distribution function H ( s , t ) = Pr( T go ≤ s , T stop ≤ t ) , for all s , t ≥ 0. ◮ The marginal distributions of H ( s , t ) are denoted as F go ( s ) = Pr( T go ≤ s , T stop < ∞ ) F stop ( t ) = Pr( T go < ∞ , T stop ≤ t ) . 51 / 75

Recommend


More recommend