normal random variable
play

Normal Random Variable X is a Normal Random Variable : X ~ N( , 2 - PowerPoint PPT Presentation

Normal Random Variable X is a Normal Random Variable : X ~ N( , 2 ) Probability Density Function (PDF): 1 2 2 2 ( x ) / f ( x ) e where x 2 E [ X ]


  1. Normal Random Variable • X is a Normal Random Variable : X ~ N( � , � 2 ) � Probability Density Function (PDF): 1 2 2 2 ( x ) / f ( x ) e where x � � � � � � � � � � 2 � � � E [ X ] � � � f ( x ) 2 Var ( X ) � � � x � Also called “Gaussian” � Note: f(x) is symmetric about � � Common for natural phenomena: heights, weights, etc. � Often results from the sum of multiple variables

  2. Carl Friedrich Gauss • Carl Friedrich Gauss (1777-1855) was a remarkably influential German mathematician • Started doing groundbreaking math as teenager � Did not invent Normal distribution, but popularized it • He looked more like Martin Sheen � Who is, of course, Charlie Sheen’s father

  3. Properties of Normal Random Variable • Let X ~ N( � , � 2 ) • Let Y = a X + b � Y ~ N( a � + b , a 2 � 2 ) � E[Y] = E[ a X + b ] = a E[X] + b = a � + b � Var(Y) = Var( a X + b ) = a 2 Var(X) = a 2 � 2 x b x b � � F ( x ) P ( Y x ) P ( aX b x ) P ( X ) F ( ) � � � � � � � � a a Y X Differentiating F Y ( x ) w.r.t. x , yields f Y ( x ) , the PDF for y : d d x b x b � � 1 f ( x ) F ( x ) F ( ) f ( ) � � � a a a dx dx Y Y X X • Special case: Z = (X – � )/ � ( a = 1/ � , b = – � / � ) � Z ~ N( a � + b , a 2 � 2 ) = N( � / � – � / � , (1/ �� � 2 � 2 ) = N(0, 1)

  4. Standard (Unit) Normal Random Variable • Z is a Standard (or Unit) Normal RV : Z ~ N(0, 1) Var(Z) = � 2 = 1 � E[Z] = � = 0 SD(Z) = � = 1 � CDF of Z, F Z ( z ) does not have closed form � We denote F Z ( z ) as � ( z ): “phi of z” z z 1 1 2 2 2 � � � � � � � ( x ) / 2 x / 2 � � � � z ) P ( Z z ) e dx e dx Φ( � 2 � 2 � � � � � � By symmetry: � ( – z ) = P( Z ≤ – z ) = P( Z ≥ z ) = 1 – � ( z ) • Use Z to compute X ~ N( � , � 2 ), where � > 0 X � � x � � x � � x � � F X ( x ) � P ( X � x ) � P ( � ) � P ( Z � ) � � ( ) � � � � � Table of � ( z ) values in textbook, p. 201 and handout

  5. Using Table of � (z) Values � (0.54) = 0.7054

  6. Get Your Gaussian On � = 3 � 2 = 16 � = 4 • X ~ N(3, 16) � What is P(X > 0)? X 3 0 3 3 � � P ( X 0 ) P ( P ( Z ) ) � � � � � � 4 4 4 3 3 1 ( ) ( ) 0 . 7734 � � � � � � 4 4 � What is P(2 < X < 5)? 2 3 X 3 5 3 1 2 � � � P ( 2 X 5 ) P ( P ( Z ) ) � � � � � � � � � 4 4 4 4 4 2 1 1 1 ( ) ( ) ( ) ( 1 ( )) 0 . 6915 ( 1 0 . 5987 ) 0 . 2902 � � � � � � � � � � � � � 4 4 2 4 � What is P(|X – 3| > 6)? 3 3 9 3 � � � P ( X 3 ) P ( X 9 ) P ( Z P ( Z ) ) � � � � � � � � 4 4 3 3 3 ( ) ( 1 ( )) 2 ( 1 ( )) 2 ( 1 0 . 9332 ) 0 . 1336 � � � � � � � � � � � 2 2 2

  7. Noisy Wires • Send voltage of 2 or -2 on wire (to denote 1 or 0) � X = voltage sent � R = voltage received = X + Y, where noise Y ~ N(0, 1) � Decode R: if (R ≥ 0.5) then 1, else 0 � What is P(error after decoding | original bit = 1)? P ( 2 Y 0 . 5 ) P ( Y 1 . 5 ) ( 1 . 5 ) 1 ( 1 . 5 ) 0 . 0668 � � � � � � � � � � � � � What is P(error after decoding | original bit = 0)? P ( 2 Y 0 . 5 ) P ( Y 2 . 5 ) 1 ( 2 . 5 ) 0 . 0062 � � � � � � � � �

  8. Normal Approximation to Binomial • X ~ Bin( n , p ) � E[X] = np Var(X) = np (1 – p ) � Poisson approx. good: n large (> 20), p small (< 0.05) � For large n: X � Y ~ N(E[X], Var(X)) = N( np , np (1 – p )) � Normal approx. good : Var(X) = np (1 – p ) ≥ 10 � � � � � � � � � � k np 0 . 5 k np 0 . 5 1 1 � � � � � � � � � � � � � � P ( X k ) P k Y k � � � � 2 2 � � np ( 1 p ) np ( 1 p ) � � � � � DeMoivre-Laplace Limit Theorem: o S n : number of successes (with prob. p ) in n independent trials � � � S np � � n � � � � � � � � � � � P a n b ( b ) ( a ) � � np ( 1 � p ) � �

  9. Comparison when n = 100, p = 0.5 P(X = k ) ) k

  10. Faulty Endorsements • 100 people placed on special diet � X = # people on diet whose cholesterol decreases � Doctor will endorse diet if X ≥ 65 � What is P(doctor endorses diet | diet has no effect)? � X ~ Bin(100, 0.5) np 50 np( 1 25 np( 1 5 � – p) � – p) � � Use Normal approximation: Y ~ N(50, 25) P ( X 65 ) P ( Y 64 . 5 ) � � � � � Y 50 64 . 5 50 � � P ( Y 64 . 5 ) P 1 ( 2 . 9 ) 0 . 0019 � � � � � � � 5 5 � Using Binomial: P ( X 65 ) 0 . 0018 � �

  11. Stanford Admissions • Stanford accepts 2480 students � Each accepted student has 68% chance of attending � X = # students who will attend. X ~ Bin(2480, 0.68) � What is P(X > 1745)? np 1686 . 4 np( 1 539 . 65 np( 1 23 . 23 � – p) � – p) � � Use Normal approximation: Y ~ N(1686.4, 539.65) P ( X 1745 ) P ( Y 1745 . 5 ) � � � � � Y 1686 . 4 1745 . 5 1686 . 4 � � P ( Y 1745 . 5 ) P 1 ( 2 . 54 ) 0 . 0055 � � � � � � � 23 . 23 23 . 23 � Using Binomial: P ( X 1745 ) 0 . 0053 � �

  12. Exponential Random Variable • X is an Exponential RV : X ~ Exp( � ) Rate: � > 0 � Probability Density Function (PDF): � � � x � � e if x 0 � � � � � � f ( x ) where x � � 0 if x 0 � 1 � E [ X ] � � f ( x ) 1 � Var ( X ) � x 2 � � Cumulative distribution function (CDF), F (X) = P( X � x ): (CDF), (X) P( ): � � x � � � F ( x ) 1 e where x 0 � Represents time until some event o Earthquake, request to web server, end cell phone contract, etc.

  13. Exponential is “Memoryless” • X = time until some event occurs � X ~ Exp( � ) � What is P(X > s + t | X > s)? P ( X s t and X s ) P ( X s t ) � � � � � P ( X s t | X s ) � � � � � P ( X s ) P ( X s ) � � ( s t ) P ( X s t ) 1 F ( s t ) e � � � � � � � t e 1 F ( t ) P ( X t ) � � � � � � � � � s P ( X s ) 1 F ( s ) e � � � � So, P ( X s t | X s ) P ( X t ) � � � � � � After initial period of time s , P(X > t | � ) for waiting another t units of time until event is same as at start � “Memoryless” = no impact from preceding period s

  14. Visits to Web Site • Say a visitor to your web leaves after X minutes � On average, visitors leave site after 5 minutes � Assume length of stay is Exponentially distributed � X ~ Exp( � = 1/5), since E[X] = 1/ � = 5 � What is P(X > 10)? 10 2 P ( X 10 ) 1 F ( 10 ) 1 ( 1 e ) e 0 . 1353 � � � � � � � � � � � � What is P(10 < X < 20)? 4 2 P ( 10 X 20 ) F ( 20 ) F ( 10 ) ( 1 e ) ( 1 e ) 0 . 1170 � � � � � � � � � � �

  15. Replacing Your Laptop • X = # hours of use until your laptop dies � On average, laptops die after 5000 hours of use � X ~ Exp( � = 1/5000), since E[X] = 1/ � = 5000 � You use your laptop 5 hours/day. � What is P(your laptop lasts 4 years)? � That is: P(X > (5)(365)(4) = 7300) 7300 / 5000 1 . 46 P ( X 7300 ) 1 F ( 7300 ) 1 ( 1 e ) e 0 . 2322 � � � � � � � � � � � Better plan ahead... especially if you are coterming: 1 . 825 P ( X 9125 ) 1 F ( 9125 ) e 0 . 1612 (5 year plan) � � � � � � 2 . 19 P ( X 10950 ) 1 F ( 10950 ) e � 0 . 1119 (6 year plan) � � � � �

  16. A Little Calculus Review • Product rule for derivatives: � ) � � � � d ( u v du v u dv • Derivative and integral of exponential: u d ( e ) du � u u e du � e u � e dx dx • Integration by parts: � � � d ( u � v ) � u � v � v � du � u � dv � � u � dv � u � v � v � du

Recommend


More recommend