Reader Comments

Post a new comment on this article

Identified risk factors vs overdispersion

Posted by GustafR on 11 Jun 2008 at 08:06 GMT

Interesting article! Is the overdispersion still present in a model for total malarial episodes that includes the factors you identified as increasing the risk? I.e, are the identified factors for "children at risk" sufficient to explain the observed overdispersion.

Regards,

Gustaf Rydevik
Statistician
Swedish Institute for Infectious Disease Control

RE: Identified risk factors vs overdispersion

tabimwangi replied to GustafR on 18 Jun 2008 at 06:01 GMT

Thank you for your interest in this paper. If I understood you correctly, you wanted to find out if putting the factors that increased the risk of total malaria episodes actually got rid of the over-dispersion. We compared the Poisson and negative binomial models in the model with and without the risk factors. We found in both that the negative binomial model fitted better than the Poisson model (p<0.001) and therefore the identified factors would not sufficiently explain the over-dispersion.