{"id":380,"date":"2017-05-12T11:10:44","date_gmt":"2017-05-12T11:10:44","guid":{"rendered":"http:\/\/jsr.isrt.ac.bd\/?post_type=article&p=380"},"modified":"2017-05-12T11:10:51","modified_gmt":"2017-05-12T11:10:51","slug":"variance-function-semi-parametric-analysis-count-data","status":"publish","type":"article","link":"http:\/\/jsr.isrt.ac.bd\/article\/variance-function-semi-parametric-analysis-count-data\/","title":{"rendered":"Variance function in semi-parametric analysis of count data"},"content":{"rendered":"
The purpose of this paper is to determine an appropriate variance function
\n(mean-variance relationship) which can be used in the semi-parametric analy-
\nsis of over-dispersed count data (for example, for analysis of count data by ex-
\ntended quasi-likelihood and double extended quasi-likelihood). We use hypothesis
\ntesting approach through a broader class of models and data analytic approach.
\nThe models considered are the three parameter negative binomial distribution
\nand the extended quasi-likelihood. Wide analysis involving tests, data analysis
\nand simulations indicate that the three parameter generalized negative binomial
\ndistribution does not improve in \ft to count data over the simpler negative bi-
\nnomial distribution. Further data analysis and simulations using the extended
\nquasi-likelihood indicate that the negative binomial variance function is
\npreferable over a simpler variance function for data with small mean and
\nsmall over-dispersion. Otherwise is a preferable variance function over the
\nnegative binomial variance function.<\/p>\n
<\/p>\n