Date:         Fri, 13 Nov 1998 01:57:42 -0600
Reply-To:     "Nichols, David" <nichols@SPSS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         "Nichols, David" <nichols@SPSS.COM>
Subject:      Re: possible to do HLM with SPSS

An earlier post asked about random effects regression models, which is
really asking the same question. Here's my response to that: No, not really.
In the special cases of balanced models where the ML and REML estimates
reduce to OLS estimates, yes, but in the more common unbalanced cases, only
the OLS solutions are available. There is no way to declare a continuous
covariate random. There are only two products that SPSS currently
distributes that provide better options for such models. The BMDP procedure
5V fits unbalanced repeated measures models using ML or REML methods. WesVar
Complex Samples 3.0 isn't designed for such models per se, but the difficult
issues introduced by such models are the dependencies among errors for
observations coming from the same units, and WesVar is designed to deal with
these problems via replication methods.
 
David Nichols
Principal Support Statistician and
Manager of Statistical Support
SPSS Inc.
 
> -----Original Message-----
> From: John S. Painter [SMTP:painter@EMAIL.UNC.EDU]
> Sent: Wednesday, November 11, 1998 9:02 AM
> To:   SPSSX-L@UGA.CC.UGA.EDU
> Subject:      possible to do HLM with SPSS
>
> Is it possible to do HLM modeling with SPSS?
>
> Thanks,
> John P