New mean of variable after adjusting for covariate for total population

[AgaWS]

Hi guys,

I want to get the mean intake of several food groups after adjusting for a continuous covariate, the total energy intake (TEI). 

I used proc glm to obtain the lsmeans for different sex as in the code below and the output:

proc glm data=test;
      class sex;
      model &foodgroup = sex TEI / solution;
      lsmeans sex / stderr pdiff diff cov out=adjmeans;
   run;
   proc print data=adjmeans;
   run;

Now I want to get the adjusted mean value for the total population and not for the genders seperatly but I couldn't find the right statement.

I would be grateful for any help

Thanks a lot! 

[gcjfernandez]

You can use  the at option with LSMEANS  to specify the value of the covariate.

[PaigeMiller]

@gcjfernandez can you provide a link to this text that you are showing?

[gcjfernandez]

The link for at option in Lsmeans:

https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.3/statug/statug_glm_details41.htm#statug.glm.gl...

This will estimate LSmeans of categorical variable levels at a known level of covariate(s)

[StatDave]

Two possible ways:

1) mean-center TEI and fit the model omitting SEX. The intercept is the estimated overall mean at the mean of TEI.

proc standard data=test m=0; var TEI; run;
proc glm; model &foodgroup = TEI / solution; run;

2) Use an ESTIMATE statement in your original GLM step - specify the mean TEI and appropriate coefficients for the levels of SEX - probably either 0.5 0.5 to represent a balanced population or the actual observed proportions of the sexes in your data. Assuming a balanced population and mean of TEI is 5.3:

estimate 'adj overall mean' intercept 1 sex .5 .5 TEI 5.3;