Conditional Likelihood for Exponential Family: Suppose that the log-likelihood for , can be written in the exponential family form l , y t s b . l , y has a decomposition of the form Also, suppose l , y t s1 t s2 b , . Note: is a linear is arbitrary The above decomposition can be achieved only if function of . The choice of nuisance parameter and the inferences regarding parameterization chosen for should be unaffected by the . ◆ The conditional likelihood of the data Y given s2 is l | s2 t s1 b , s2 , which is independent of the nuisance parameter and may be used for inferences regarding . Example : Y1 ~ P1 , Y2 ~ P 2 are independent. Suppose 2 log 2 log 1 is the parameter 1 log 1 of interest and 1 log 1 is the nuisance parameter. Then, the log-likelihood is l , 1 log e 1 2 1y1 2y 2 1 2 y1 log 1 y 2 log 2 2 1 1 y1 log 1 y 2 log 1 1 y 2 log 1 log 2 e 1 1 e y1 y 2 1 y 2 s1 s2 1 b , 1 s1 y2 , s2 y1 y2 , b , 1 e 1 1 e Then, the conditional distribution of Y1 , Y2 given s2 Y1 Y2 m 1 B m , is . Thus, 1 2 l | s 2 m 1 y1 log 2 1 2 y log 2 2 1 1 y1 log 2 1 1 y log 2 2 1 1 y 2 log 2 1 . 2 log 2 1 1 1 2 1 y1 y 2 log 1 y 2 1 exp 2 1 1 s1 b , s 2 2 1 b , s s log 2 2 1 exp . where Note: We can choose the other nuisance parameters, for example, 2 log 2 , 3 log 1 2 , 4 log 1 log 2 . Thus, the log-likelihood is l , 2 1 2 y1 log 1 y 2 log 2 y1 y1 y 2 2 e 2 1 e l , 4 y y2 y1 y2 1 2 4 2 2e 4 2 cosh 2 No matter what parameterization is chosen for the nuisance parameter, the sufficient statistic S 2 Y1 Y2 has the same form. Example : Y1 ~ B1, 1 , Y2 ~ B1, 2 are independent. Suppose 2 1 log 1 1 1 2 log interest and 2 1 2 log Then, the log-likelihood is 3 is the parameter of is the nuisance parameter. l , 2y 1 2 1 y y1 log 1 1 y1 log 1 1 y2 log 2 1 y2 log 1 2 log 1y1 1 1 1 y1 2 2 2 1 log y y1 log 2 1 1 2 1 log 1 1 log 1 2 2 1 log y y1 log 1 1 1 2 1 2 y1 y2 log log 1 1 log 1 2 1 2 y1 y1 y2 log 1 1 log 1 2 s1 s2 b , s1 y1 , s2 y1 y2 , b , log 1 1 log 1 2 . Then, the conditional likelihood can be obtained based on the conditional distribution of Y1 , Y2 given s2 Y1 Y2 m , which is the hypergeometric distribution as 4 1 2 .