Subscribe Bookmark RSS Feed

How to calculate scale estimate in GLS

angela

Community Trekker

Joined:

Sep 6, 2016

Scaled estimated is available for ordinary least square analysis. But I can't find this information when I use GLS? Can anyone explain this? Has this something to do with assumption of residual normal distribution?

4 REPLIES
jiancao

Staff

Joined:

Jul 7, 2014

GLM doesn't provide the scaled estimates. You could use the JMP formula editor to create the scaled factors and then run the GLM.

Scaling doesn't have to do with the distribution assumptions; by scaling it puts factors on an equal footing to enable the comparison of the estimated coefficients across different factors originally in different measurement units.

If you use JMP Pro, in its Generalized Regression,the  parameter estimate report by default shows the results for both original and scaled factors. Note that GenReg uses a slightly different scaling method--it subtracts the mean and then is divided by the standard deviation where as in OLS after the mean is subtracted it is divided by range/2.

angela

Community Trekker

Joined:

Sep 6, 2016

Thanks for explanation. I use JMP 12. Do you know where to find effect size for the report using REML method?

jiancao

Staff

Joined:

Jul 7, 2014

Currently there is a JMP Add-in for effect size from OLS modes Calculate Effect Sizes Add-in

Random effects model are not supported.

angela

Community Trekker

Joined:

Sep 6, 2016

Thanks for the add-in. I am wondering if this can be used to estimate effect size for fixed effects in the report for a split-plot design, which does have random effects (whole and subplots). I am so used to use scaled estimate for effect size evaluation in ordinary least square analysis. Now I can't find similar information in the regression report for a split plot design.