glme reference group treatment

["Ali & Hind Lazrak" <alazrak@telus.net>]

Hello

 

As part of my thesis (Master's level), I need to predict the use of a device. The data that was gathered is a yes/no use self reported answer.

I fitted a multiple logistic regression mixed model with random intercepts to model a binary outcome the glme function from the correlatedData library (Splus version 8).

My data has the following structure

14 companies : outer level

10,095 subjects: inner level

59,000 measurements of the binary outcome.

A GEE model was first fitted to the data to obtain the starting values that were entered in the final model for parameters estimation using Restricted Penalized Quasi Likelihood method.

After review of the first draft of the paper I'm working on, my supervisors asked me to change the reference group of two of the fixed effects variables (gender and race) to the group having the larger number of observations.

Right now, I have female and Chinese.

To do what has been asked of me I need to switch the reference groups to Male and White.

Race is coded: Chinese, Indian, White

Gender is coded: Female, Male

I have set the options as follow:

options(contrasts=c(factor="contr.treatment",ordered="contr.poly))

I used the GUI interface and recoded Female as Women and White as AWhite

However I still have an output where Women and Chinese are the reference group of Gender and Race respectively.

 

I truly hope that my question was clear and that you can provide me with a solution/ explanation as to why this is happening...

 

Many thanks in advance

 

Hind