3.8 Functions of random variables 3.7, 3.9, 3.11 Multiple random variables (discrete) Prof. Tesler Math 186 Winter 2019 Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 1 / 39
3.8 One random variable as a function of another random variable (Discrete) Let X = roll of a biased die, Y = 10 X + 2 , and Z = ( X − 3 ) 2 . z = ( x − 3 ) 2 p X ( x ) y = 10 x + 2 p Y ( y ) p Z ( z ) x 1 q 1 12 q 1 4 2 q 2 22 q 2 1 p Z ( 0 ) = q 3 3 q 3 32 q 3 0 p Z ( 1 ) = q 2 + q 4 4 q 4 42 q 4 1 p Z ( 4 ) = q 1 + q 5 5 q 5 52 4 q 5 p Z ( 9 ) = q 6 6 62 9 q 6 q 6 For W = g ( X ) , the pdf p W ( w ) is the sum of p X ( x ) over all possible inverses x = g − 1 ( w ) , or p W ( w ) = 0 if there are no inverses: � p W ( w ) = p X ( x ) x : g ( x )= w Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 2 / 39
Function of a Discrete Random Variable Let X = roll of a biased die, Y = 10 X + 2 , and Z = ( X − 3 ) 2 . z = ( x − 3 ) 2 p X ( x ) y = 10 x + 2 p Y ( y ) p Z ( z ) x 1 q 1 12 q 1 4 2 q 2 22 q 2 1 p Z ( 0 ) = q 3 3 q 3 32 q 3 0 p Z ( 1 ) = q 2 + q 4 4 q 4 42 q 4 1 p Z ( 4 ) = q 1 + q 5 5 q 5 52 q 5 4 p Z ( 9 ) = q 6 6 q 6 62 q 6 9 For Y = aX + b with a � 0 , the unique inverse is X = Y − b a , so � y − b � p Y ( y ) = p X a Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 3 / 39
Function of a Discrete Random Variable Let Y = X 2 . Inverses of y = x 2 are ± √ y if y > 0 ; x = if y = 0 ; 0 none if y < 0 . P ( X = √ y ) + P ( X = − √ y ) if y > 0 ; P ( Y = y ) = P ( X 2 = y ) = P ( X = 0 ) if y = 0 ; if y < 0 . 0 Use pdf notation to express pdf of Y in terms of pdf of X : p X ( √ y ) + p X (− √ y ) if y > 0 ; p Y ( y ) = p X ( 0 ) if y = 0 ; if y < 0 . 0 Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 4 / 39
Function of a Continuous Random Variable Let U be uniform on the real interval [ 1 , 7 ] , so � if 1 � u � 7 (in real numbers); 1 / 6 f U ( u ) = otherwise 0 Let V = 2 U . This is uniform on [ 2 , 14 ] , so � 1 / 12 if 2 � v � 14 (in real numbers); f V ( v ) = otherwise 0 Even though there’s just one inverse for each value, the probabilities of corresponding values are different! Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 5 / 39
Continuous Random Variables: Computing f Y ( y ) from f X ( x ) when Y = g ( X ) Procedure Let Y = g ( X ) . Compute F X ( x ) . Compute F Y ( y ) = P ( Y � y ) = P ( g ( X ) � y ) = · · · Details depend on the function. Typically it is expressed in terms of F X ( x ) at various values of x (the ones where g ( x ) = y ). Compute f Y ( y ) = d dyF Y ( y ) . Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 6 / 39
Example with one-to-one function f X ( x ) = 8 e − 8 x Let for x � 0 . It’s a valid pdf since it’s � 0 and the total probability is 1: � ∞ � ∞ 8 e − 8 x dx = 8 e − 8 x ∞ � � f X ( x ) dx = � − 8 � − ∞ 0 x = 0 = −( e − ∞ − e 0 ) = −( 0 − 1 ) = 1 The CDF for x � 0 is � x � x x 8 e − 8 t dt = 8 e − 8 t � � F X ( x ) = f X ( t ) dt = � − 8 � − ∞ 0 t = 0 = −( e − 8 x − e 0 ) = 1 − e − 8 x while for x < 0 , the CDF is F X ( x ) = 0 . Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 7 / 39
Example with one-to-one function � � 8 e − 8 x 1 − e − 8 x if x � 0 ; if x � 0 ; pdf: f X ( x ) = cdf: F X ( x ) = if x < 0 if x < 0 0 0 We’ll compute the pdf of Y = 10 X + 2 . First, we convert the two cases x � 0 and x < 0 to y . Note x � 0 gives 10 x � 0 , so 10 x + 2 � 2 , so y � 2 . Similarly, x < 0 gives y < 2 . Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 8 / 39
Example with one-to-one function � � 8 e − 8 x 1 − e − 8 x if x � 0 ; if x � 0 ; f X ( x ) = F X ( x ) = Y = 10 X + 2 if x < 0 if x < 0 0 0 Wrong Way: f Y ( y ) = f X ( g − 1 ( y )) as in the discrete case Try f Y ( y ) = f X (( y − 2 ) / 10 ) : � 8 e − 8 ( y − 2 ) / 10 if y � 2 ; f Y ( y ) = if y < 2 . 0 Compute the total probability: � ∞ � ∞ ∞ 8 e − 8 ( y − 2 ) / 10 dy = 8 e − 8 ( y − 2 ) / 10 � � f Y ( y ) dy = � − 8 / 10 � − ∞ 2 y = 2 = − 10 ( e − ∞ − e 0 ) = − 10 ( 0 − 1 ) = 10 � 1 The total probability is not 1, so this is not a valid pdf. Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 9 / 39
Example with one-to-one function � � 8 e − 8 x 1 − e − 8 x if x � 0 ; if x � 0 ; f X ( x ) = F X ( x ) = Y = 10 X + 2 if x < 0 if x < 0 0 0 Right way: Compute CDF F Y ( y ) and differentiate f Y ( y ) = F Y ′ ( y ) Compute CDF F Y ( y ) : F Y ( y ) = P ( Y � y ) = P ( 10 X + 2 � y ) � 1 − e − 8 ( y − 2 ) / 10 � � � y − 2 � X � y − 2 if y � 2 ; = P = F X = 10 10 if y < 2 . 0 Differentiate to get PDF f Y ( y ) : � 10 e − 8 ( y − 2 ) / 10 8 if y > 2 ; ′ ( y ) = f Y ( y ) = F Y if y < 2 . 0 Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 10 / 39
Y = aX + b in general where a , b are constants and a � 0 In general, for a continuous random variable X , the pdfs of X and Y = aX + b are related by � y − b � f Y ( y ) = 1 | a | f X a Whereas if X is a discrete random variable, then � y − b � p Y ( y ) = p X a Note the scaling factor 1 / | a | for the continuous case, but not for the discrete case. Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 11 / 39
Example with general function Y = g ( X ) Compute CDF and PDF of Y = X 2 where X is a continuous R.V. When y � 0 , F Y ( y ) = P ( Y � y ) = P ( X 2 � y ) = 0 f Y ( y ) = d dyF Y ( y ) = d dy 0 = 0 When y > 0 , = P ( Y � y ) = P ( X 2 � y ) F Y ( y ) = P (− √ y � X � √ y ) = F X ( √ y ) − F X (− √ y ) = d F X ( √ y ) − F X (− √ y ) � � f Y ( y ) dy ′ ( √ y ) ′ (− √ y ) � � � � 1 1 = F X − F X − 2 √ y 2 √ y = f X ( √ y ) + f X (− √ y ) 2 √ y Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 12 / 39
3.7, 3.9, 3.11 Multiple random variables (discrete) Multinomial distribution Consider a biased 6-sided die where q i is the probability of rolling i for i = 1 , 2 , . . . , 6 . (6 sides is an example, it could be any # sides.) Each q i is between 0 and 1 , and q 1 + · · · + q 6 = 1 . The probability of a sequence of independent rolls is 6 � P ( 1131326 ) = q 1 q 1 q 3 q 1 q 3 q 2 q 6 = q 13 q 2 q 32 q 6 = q i # i ’s i = 1 Roll the die n times ( n = 0 , 1 , 2 , 3 , . . .). Let X 1 be the number of 1’s, X 2 be the number of 2’s, etc. p X 1 , X 2 ,..., X 6 ( k 1 , k 2 , . . . , k 6 ) = P ( X 1 = k 1 , X 2 = k 2 , . . . , X 6 = k 6 ) q 1 k 1 q 2 k 2 . . . q 6 k 6 n � � k 1 , k 2 ,..., k 6 if k 1 , . . . , k 6 are integers � 0 adding up to n ; = otherwise. 0 Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 13 / 39
Genetics example Consider a TtRR × TtRr cross of pea plants: Genotype Prob. Punnett Square 1 / 8 TTRR TR ( 1 / 2 ) tR ( 1 / 2 ) 2 / 8 = 1 / 4 TtRR TR ( 1 / 4 ) TTRR ( 1 / 8 ) TtRR ( 1 / 8 ) 1 / 8 TTRr Tr ( 1 / 4 ) TTRr ( 1 / 8 ) TtRr ( 1 / 8 ) 2 / 8 = 1 / 4 TtRr tR ( 1 / 4 ) TtRR ( 1 / 8 ) ttRR ( 1 / 8 ) 1 / 8 ttRR tr ( 1 / 4 ) TtRr ( 1 / 8 ) ttRr ( 1 / 8 ) 1 / 8 ttRr Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 14 / 39
Genetics example If there are 27 offspring, what is the probability that 9 offspring have genotype TTRR, 2 have genotype TtRR, 3 have genotype TTRr, 5 have genotype TtRr, 7 have genotype ttRR, and 1 has genotype ttRr? Use the multinomial distribution: Genotype Probability Frequency 1/8 TTRR 9 1/4 TtRR 2 1/8 TTRr 3 1/4 TtRr 5 1/8 ttRR 7 1/8 ttRr 1 Total 1 27 � 9 � 1 � 2 � 1 � 3 � 1 � 5 � 1 � 7 � 1 � 1 � 1 27 ! ≈ 2 . 19 · 10 − 7 P = 9 ! 2 ! 3 ! 5 ! 7 ! 1 ! 8 4 8 4 8 8 Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 15 / 39
Genetics example If there are 25 offspring, what is the probability that 9 offspring have genotype TTRR, 2 have genotype TtRR, 3 have genotype TTRr, 5 have genotype TtRr, 7 have genotype ttRR, and 1 has genotype ttRr? P = 0 because the numbers 9 , 2 , 3 , 5 , 7 , 1 do not add up to 25 . Prof. Tesler Ch. 3. Functions of random vars. (discrete) Math 186 / Winter 2019 16 / 39
Recommend
More recommend