Multivariate Probability Distributions Multivariate Random Variables • In many settings, we are interested in 2 or more characteristics observed in experiments • Often used to study the relationship among characteristics and the prediction of one based on the other(s) • Three types of distributions: – Joint: Distribution of outcomes across all combinations of variables levels – Marginal: Distribution of outcomes for a single variable – Conditional: Distribution of outcomes for a single variable, given the level(s) of the other variable(s) Joint Distribution Discrete Case (Probabili ty Mass Function) : p( y1 , y2 ) PY1 y1 , Y2 y2 0 p( y , y ) 1 1 2 all y1 all y 2 F ( y1 , y2 ) PY1 y1 , Y2 y2 y1 y2 p( y , y ) t1 t 2 1 2 Continuous Case (Probabili ty Density Function) : f ( y1 , y2 ) 0 f ( y1 , y2 )dy2 dy1 1 F ( y1 , y2 ) PY1 y1 , Y2 y2 y1 y2 f (t1 , t 2 )dt 2 dt1 Generalize s to any number of Random Variables Marginal Distributions Discrete Case : p1 ( y1 ) p( y , y ) 1 2 all y 2 p2 ( y 2 ) p( y , y ) 1 2 all y1 Continuous Case : f1 ( y1 ) f ( y1 , y2 )dy2 f 2 ( y2 ) f ( y1 , y2 )dy1 Generalize s to any number of Random Variables (sum or integrate over all other vari ables) Conditional Distributions • Describes the behavior of one variable, given level(s) of other variable(s) Discrete Case: p ( y1 | y2 ) P Y1 y1 | Y2 y2 p ( y2 | y1 ) P Y2 y2 | Y1 y1 Continuous Case: f ( y1 , y2 ) f ( y1 | y2 ) f 2 ( y2 ) f ( y1 , y2 ) f ( y2 | y1 ) f1 ( y1 ) p ( y1 , y2 ) p2 ( y2 ) p ( y1 , y2 ) p1 ( y1 ) p( y 1 | y2 ) 1 y2 s.t. p2 ( y2 ) 0 all y1 p( y 2 | y1 ) 1 y1 s.t. p1 ( y1 ) 0 all y2 f ( y1 | y2 )dy1 1 y2 s.t. f 2 ( y2 ) 0 f ( y2 | y1 )dy2 1 y1 s.t. f1 ( y1 ) 0 Expectations Discrete Case : E g (Y1 , Y2 ) g ( y , y ) p( y , y ) 1 2 1 2 all y1 all y 2 E Y1 1 y p( y , y ) y p( y , y ) y p ( y ) 1 1 2 all y1 all y 2 V (Y1 ) 12 1 all y1 (y 1 1 1 1 1 all y1 (y 1 ) 2 p ( y1 , y2 ) all y1 all y 2 2 all y 2 1 1 ) 2 p1 ( y1 ) all y1 Continuous Case : E g (Y1 , Y2 ) E Y1 1 V (Y1 ) 2 1 g ( y1 , y2 ) f ( y1 , y2 )dy2 dy1 y1 f ( y1 , y2 )dy2 dy1 y1 f ( y1 , y2 )dy2 dy1 y1 f1 ( y1 )dy1 ( y1 1 ) f ( y1 , y2 )dy2 dy1 ( y1 1 ) 2 f1 ( y1 )dy1 2 Covariance of Y1 , Y2 : COV (Y1 , Y2 ) E Y1 1 Y2 2 E Y1Y2 Y1 2 1Y2 1 2 E Y1Y2 2 E Y1 1 E Y2 1 2 E Y1Y2 1 2 Expectations of Linear Functions Y1 ,..., Yn Random Variables with E (Yi ) i X 1 ,..., X m Random Variables with E ( X j ) j n m i 1 j 1 U1 aiYi U 2 b j X j {ai },{b j } constants a1 y1 ... an yn f ( y1 ,..., yn )dyn ...dy1 E (U1 ) ... a1 ... y1 f ( y1 ,..., yn )dyn ...dy1 ... an ... yn f ( y1 ,..., yn )dyn ...dy1 a1 E (Y1 ) ... an E (Yn ) n ai i i 1 Variances of Linear Functions Y1 ,..., Yn Random Variables with E (Yi ) i X 1 ,..., X m Random Variables with E ( X j ) j n m i 1 j 1 U1 aiYi U 2 b j X j {ai },{b j } constants 2 n n 2 V (U1 ) E (U1 E (U1 )) E aiYi ai i i 1 i 1 2 n E ai (Yi i ) i 1 n 1 n n 2 2 E ai (Yi i ) 2 ai (Yi i )ai ' (Yi ' i ' ) i 1 i 'i 1 i 1 n n 1 a E (Yi i ) 2 i 1 2 i 2 a V (Yi ) 2 i 1 2 i a a E(Y )(Y i 1 i 'i 1 n 1 n n i i' i n a a COV (Y , Y ) i 1 i 'i 1 i i' i i' i i' i ' ) Covariance of Two Linear Functions Y1 ,..., Yn Random Variables with E (Yi ) i X 1 ,..., X m Random Variables with E ( X j ) j n m i 1 j 1 U1 aiYi U 2 b j X j {ai },{b j } constants m n COV (U1 , U 2 ) COV aiYi , b j X j j 1 i 1 n m n m E aiYi ai i b j X j b j j i 1 j 1 j 1 i 1 m n E ai (Yi i ) b j ( X j j ) j 1 i 1 n m n m ai b j E (Yi i )( X j j ) ai b j COV (Yi , X j ) i 1 j 1 i 1 j 1 Multinomial Distribution • Extension of Binomial Distribution to experiments where each trial can end in exactly one of k categories • n independent trials • Probability a trial results in category i is pi • Yi is the number of trials resulting in category I • p1+…+pk = 1 • Y1+…+Yk = n Multinomial Distribution p y1 ,..., yk PY1 y1 ,..., Yk yk n! yk y1 p1 ... pk y1!... yk ! k k i 1 i 1 yi n, pi 1, yi 0, pi 0 n! pi ( yi ) piyi (1 pi ) n yi yi !(n yi )! yi 0,1,.., n (Yi has a marginal binomial distributi on) E (Yi ) npi V (Yi ) npi (1 pi ) Multinomial Distribution Covariance of Y j , Y j ' : 1 if trial i results in category j Ui 0 otherwise E (U i ) 1( p j ) 0(1 p j ) p j 1 if trial i results in category j ' Vi 0 otherwise E (Vi ) p j ' E (U iVi ) 1(0) 0(1) 0 (Each tria l can result in only one category) COV (U i , Vi ) E (U iVi ) E (U i ) E (Vi ) 0 p j p j ' p j p j ' COV (U i , Vi ' ) 0 i i ' by independen ce n Y j U i i 1 n Y j ' Vi i 1 n n n n COV Y j , Y j ' COV U i , Vi COV U i ,Vi ' i 1 i 1 i 1 i '1 n n COV U i , Vi COV U i , Vi ' np j p j ' i 1 i 1 i ' i Conditional Expectations Discrete Case : E Y1 | y2 E Y1 | Y2 y2 V Y1 | y2 V Y1 | Y2 y2 y p( y 1 all y1 y 1 1 | y2 ) E Y1 | y2 p ( y1 | y2 ) 2 all y1 Continuous Case : E Y1 | y2 E Y1 | Y2 y2 y1 f ( y1 | y2 )dy1 V Y1 | y2 V Y1 | Y2 y2 y1 EY1 | y2 2 f ( y1 | y2 )dy1 When E[Y1|y2] is a function of y2, function is called the regression of Y1 on Y2 Unconditional and Conditional Mean E Y1 y1 f1 ( y1 )dy1 y1 f ( y1 , y2 )dy2 dy1 y1 f ( y1 | y2 ) f 2 ( y2 )dy2 dy1 y1 f ( y1 | y2 )dy1 f 2 ( y2 )dy2 E Y1 | y2 f 2 ( y2 )dy2 EY2 E Y1 | Y2 Unconditional and Conditional Variance V Y1 | Y2 E Y | Y2 E Y1 | Y2 2 1 2 EY2 V Y1 | Y2 EY2 E Y | Y2 E Y1 | Y2 2 1 2 E Y E E Y | Y E Y E Y E E Y | Y E Y V Y E E Y | Y E E Y | Y EY2 E Y | Y2 EY2 E Y1 | Y2 2 2 1 2 1 2 Y2 1 2 2 2 1 2 1 Y2 1 2 2 1 2 1 Y2 1 2 V Y1 VY2 E Y1 | Y2 2 Y2 V (Y1 ) E[V (Y1 | Y2 )] V [ E (Y1 | Y2 )] 1 2 Compounding • Some situations in theory and in practice have a model where a parameter is a random variable • Defect Rate (P) varies from day to day, and we count the number of sampled defectives each day (Y) – Pi ~Beta(a,b) Yi |Pi ~Bin(n,Pi) • Numbers of customers arriving at store (A) varies from day to day, and we may measure the total sales (Y) each day – Ai ~ Poisson(l) Yi|Ai ~ Bin(Ai,p)