topic Re: ODS Table Output for PROC GLM and LSMEANS in Statistical Procedures

ODS Table Output for PROC GLM and LSMEANS

LEN1 — Wed, 27 Jul 2016 15:48:00 GMT

May you please assist, from the below syntax for an augumented design

Proc glm;

Class blk X C;

Model NP NE MOI FW GW GLS PLS HM1 MSV GDS = blk C X(C);

RANDOM BLK/TEST;

lsMEANS C X(C)/STDERR;

RUN;

i am getting no-est for the LS mean

Re: ODS Table Output for PROC GLM and LSMEANS

Ksharp — Thu, 28 Jul 2016 03:09:11 GMT


Try :

Model NP NE MOI FW GW GLS PLS HM1 MSV GDS =  blk C X(C) / solution ;

Re: ODS Table Output for PROC GLM and LSMEANS

LEN1 — Thu, 28 Jul 2016 12:15:18 GMT

Hi Esteemed

The output that came out has anova tables that do not have F-values and P>F. There is also no values for Coeff var and root MSE. For the estimates there are also no value for sterror, t Value and Pr >/t/. The LSmean is still Non-est. May you please assist again.

Re: ODS Table Output for PROC GLM and LSMEANS

SteveDenham — Thu, 28 Jul 2016 14:54:27 GMT

The non-estimability is almost certainly due to empty cells. You are fitting C and X*C, and if every level of C is not "connected" for every level of X, then C cannot be estimated at the levels where connection doesn't happen. See the documentation regarding estimable functions for more depth on this. As a way to check this, add this statement to your GLM:

MEANS X(C);

and look for missing cells.

Steve Denham