Date: Tue, 3 Feb 1998 08:59:46 -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: var comp revisited
You're asking VARCOMP to estimate a model with about 72000
nonredundant (and 140,140 including redundant) parameters.
By my (admittedly somewhat imprecise) calculations, it would
take several GB of memory just to hold the design matrix.
Principal Support Statistician and
Manager of Statistical Support
>From: Josh Sacco[SMTP:josh@PILOT.MSU.EDU]
>Sent: Monday, February 02, 1998 5:55 PM
>Subject: Re: var comp revisited
>At 04:23 PM 2/2/98 -0600, Nichols, David wrote:
>>How many levels of the factors are there?
>'E' is the subject number (around 1,000 subjects), 'D' has 4 levels, 'R'
>has 3, 'P' has 6
>>Principal Support Statistician and
>>Manager of Statistical Support
>>>From: Josh Sacco[SMTP:josh@PILOT.MSU.EDU]
>>>Sent: Saturday, January 31, 1998 7:11 PM
>>>Subject: var comp revisited
>>>I'm using the following syntax:
>>> y BY d e p r
>>> /RANDOM = d e p r
>>> /METHOD = REML
>>> /CRITERIA = ITERATE(50)
>>> /CRITERIA = CONVERGE(1.0E-8)
>>> /PRINT = HISTORY (1)
>>> /design=E D R P E*D E*R(p) E*P D*R(p) D*P E*D*R(p) E*D*P.
>>>and am getting the following error:
>>>The large number of columns in the design matrix will cause integer
>>>The large number of columns may be due to too many levels in one or more
>>>factors, or to higher-order interactions among factors with many levels. It
>>>may also be due to too many factors.
>>>This command not executed.
>>>any ideas? using values of up to 175MB for the above memory setting didn't