topic Re: Means between two time points got doubled but still not significant in Statistical Procedures

Means between two time points got doubled but still not significant

Bhupinder — Fri, 11 Mar 2011 19:41:20 GMT

Hi,

The means between two time points got doubled but still differnces is not coming out significant in MIXED model. Any reason?

Thanks
Bhupinder

Re: Means between two time points got doubled but still not significant

Doc_Duke — Fri, 11 Mar 2011 20:19:08 GMT

Things to look for:

-- outliers
-- small sample sizes
-- high residual variability

Any one of those, or a combination, could cause the situation you have observed. Data plots are often helpful.

Re: Means between two time points got doubled but still not significant

Bhupinder — Fri, 11 Mar 2011 20:34:15 GMT

Duke,

Thanks for another suggestion. I have found high residual varibility. I have tried to incorporate different model parameters and when looks at the residual estimates from covariance parameter estimates, it is huge compared to other parameter's estimates. Any remedy to deal with it?

Re: Means between two time points got doubled but still not significant

statsplank — Mon, 14 Mar 2011 12:30:51 GMT

Hello Bhupinder,

I'd suggest checking if there is any pattern in the residual vs. predicted plot. This might suggest if you need to transform the outcome variable.