EMMEANS for LMATRIX planned comparisons in GLM

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

EMMEANS for LMATRIX planned comparisons in GLM

Veitch, Jennifer
We use GLM to run MANCOVA models with 1-df planned comparisons. We use
LMATRIX to specify the planned comparisons and to obtain the
multivariate test. We need to obtain the adjusted marginal means for the
specific contrast, but we are having a problem obtaining exactly what we
are looking for.  Usually we contrast two groups (experimental
conditions) and use EMMEANS to obtain the estimated marginal means, but
in this case we are doing a weighted contrast of  one group against two
others, while leaving one other out of this test entirely.

That is, we have two dependent factors: TIME and TRIES each measured
during the completion of a task. The independent variable is GROUP, with
four levels, and the covariate is AGE.   We need the EMMEANS associated
with the specified LMATRIX (two of them). Below is the syntax we have
used.

GLM     TRIES
        TIME
BY GROUP WITH AGE              
/LMATRIX                =     "1 vs 4"
                                        GROUP 1 0 0 -1
        /EMMEANS                =     TABLE (GROUP)
        /PRINT                  =     DESCRIPTIVES
                                        ETASQ
                                        TEST (LMATRIX).

It's easy to get the EMMEANS for this LMATRIX, but we are stumped on how
to get the adjusted means for the contrast described in the command that
follows:

        /LMATRIX                =     "2+3 vs 4"
                                        GROUP 0 1/2 1/2 -1


We need to know the estimated means that are contrasted in this LMATRIX
- the combination of groups 2 and 3, against group 4. Is there a way to
obtain this?


---------------------------------------------------------------------
Jennifer A. Veitch, Ph.D.
Senior Research Officer, Indoor Environment Research Program

t +1-613-993-9671 | f +1-613-954-3733 | [hidden email]

NRC Institute for Research in Construction (NRC-IRC) | 1200 Montreal
Road, M-24, Ottawa, Ontario K1A 0R6

Lighting Research:  http://irc.nrc-cnrc.gc.ca/ie/lighting/index_e.html
Cost-Effective Open-Plan Environments:
http://irc.nrc-cnrc.gc.ca/ie/cope/index_e.html
Healthy Buildings: http://irc.nrc-cnrc.gc.ca/health/index_e.html

Government of Canada

====================To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD