RS,
Here's code to fit a binary logistic regression model with two fixed
effects (and a fixed grand intercept), along with a random intercept.
GENLINMIXED
/FIELDS TARGET=y
/TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=LOGIT
/FIXED EFFECTS=x1 x2 USE_INTERCEPT=TRUE
/BUILD_OPTIONS TARGET_CATEGORY_ORDER=DESCENDING
/RANDOM USE_INTERCEPT=TRUE SUBJECTS=subject
COVARIANCE_TYPE=VARIANCE_COMPONENTS.
where
"y" = binary dependent variable
"x1" and "x2" = continuous fixed effects variables
"subject" = subject identification variable
HTH,
Ryan
On Tue, May 22, 2012 at 3:48 PM, Rieza Soelaeman <
[hidden email]> wrote:
> Dear SPSSX-Lers,
> I'm new to SPSS. I'm analyzing data collected using multistage cluster
> sampling design, as such the clusters and the sampling weights must be
> considered in the analysis. I want to fit a multilevel (mixed) logit model
> to control for clustering. I was thinking about doing a simple random
> intercept model, because in my case, fitting a random slopes model doesn't
> really make sense. I tried to Google syntax for this kind of modeling, but
> did not have any luck.
>
> Wondering if anyone on this listserv has done such an analysis, or could
> point me to a resource/samples for doing this kind of thing in SPSS.
>
> The model I'm thinking of:
>
> Ever delivered at facility [yes/no] = FIXED PORTION: [individual level
> covariates] + [community level covariates] + intercept_ij + RANDOM PORTION:
> intercept_j
>
> Controlling for non-self weighting design.
>
> any ideas?
>
> Thanks,
> RS
>
>
=====================
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
Locked
