topic Re: Effect size in GLIMMIX in Statistical Procedures

Effect size in GLIMMIX

cedric5 — Thu, 19 Mar 2020 18:27:07 GMT

Hello,

I ran a mixed model using GLIMMIX with 3 fixed effects (category) and their interaction and 2 covariates.

One the fixed effect (population) which has 4 populations is significant. Using LSMEANS I got multiple comparisons and it showed that 2 populations are different from the 2 others. I have been asked to give an effect size of the difference.

Can I get it from GLIMMIX?

Thanks for you help

Cedric

Re: Effect size in GLIMMIX

PaigeMiller — Thu, 19 Mar 2020 18:39:43 GMT

The "effect size" can be computed from the difference between the LSMEAN for Y for that group compared to the overall average for Y.

Re: Effect size in GLIMMIX

cedric5 — Thu, 19 Mar 2020 18:52:01 GMT

Thank you Paige. By Y you mean the estimate for the group?

Re: Effect size in GLIMMIX

PaigeMiller — Thu, 19 Mar 2020 19:09:08 GMT

Why don't you show us the output so we can all be talking about concrete numbers?

Re: Effect size in GLIMMIX

cedric5 — Thu, 19 Mar 2020 19:22:25 GMT

Thank you again
Here is the output of the LSMEANS

Location Least Squares Means
Location Estimate Standard Error DF t Value Pr > |t| Alpha Lower Upper
AK 5.1572 0.04854 77.17 106.24 <.0001 0.05 5.0605 5.2539
NY 4.8595 0.05958 59.38 81.57 <.0001 0.05 4.7403 4.9787
TN 4.8979 0.04920 91.59 99.56 <.0001 0.05 4.8002 4.9956
WY 5.1065 0.06950 102.3 73.48 <.0001 0.05 4.9687 5.2444

Re: Effect size in GLIMMIX

PaigeMiller — Thu, 19 Mar 2020 19:25:41 GMT

In the future, would you please click on the </> icon and paste your text output into the window that appears to preserve the formatting and make your results easier to read?

The "effect size" for AK is 5.1572 – (overall mean of Y)

One could also define "effect size" of NY to AK, which would be 5.1572 – 4.8595

Re: Effect size in GLIMMIX

cedric5 — Thu, 19 Mar 2020 19:29:15 GMT

Thank you so much!
Sorry about the formatting problem it is the first time I post.