The procedure of poisson regression model using lower respiratory illness in infants / Zuraira Libasin, Suryaefiza Karjanto and Shamsunarnie Mohamed Zukri
This study considers an analysis using a Poisson regression model where the response outcome is a count, with large outcomes being rare events. Estimates of the parameters are obtained by using the maximum likelihood estimates. Inferences about the regression parameters are based on Wald test and li...
Saved in:
Main Authors: | , , |
---|---|
Format: | Book |
Published: |
Universiti Teknologi MARA Pulau Pinang,
2014.
|
Subjects: | |
Online Access: | Link Metadata |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | This study considers an analysis using a Poisson regression model where the response outcome is a count, with large outcomes being rare events. Estimates of the parameters are obtained by using the maximum likelihood estimates. Inferences about the regression parameters are based on Wald test and likelihood ratio test. In the model building process, the stepwise selection method were used to determine important predictor variables, diagnostic tools were used in detecting multicollinearity, non-constant variance, outliers, and also analysis of residual were used to measure the goodness fit of the model. Applications of these methods are illustrated by employing a study from LaVange, Keyes, Koch, and Margolis (1994) where a case study of lower respiratory illness data in infants which took repeated observations of infants over one year. Six explanatory variables involve the number of weeks during that year for which the child is considered to be at risk, crowded conditions occur in the household, family's socioeconomic status, race, passive smoking, and age group. We found that the explanatory variables which contribute significantly are passive smoking and crowding. Social economic status and race do not appear to be influential, and neither does age group.The value of R2 is 0.0562 which indicate that about 5.62% from the total variation can be explained by the Poisson regression model. This number does not give a better result since the variance is non-constant. It simply means the existence of overdispersion. |
---|---|
Item Description: | https://ir.uitm.edu.my/id/eprint/10663/1/Zuraira%20Libasin.pdf |