Mixed question

[Maguin, Eugene]

All,
Using Mixed V12, a model such as

Mixed y with time/fixed time/random intercept time | subject(id) cov(ID)

reports a residual variance and an intercept + time variance.

V12+ may be different (and I haven't checked) but I would have expected to
see separate estimates for intercept variance, slope (time) varince and
intercept-slope covariance (as in an LGM). My reading of the documentation
indicates to me that printing G should show the desired estimates but it
doesn't. I either have to be doing something wrong or don't understand how
mixed works. Help please.

Thanks, Gene Maguin

=====================
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

[Ryan]

Gene,

The covariance structure you specified assumes no correlation between random components. I suggest you consider specifying something such as an unstructured (UN) covariance matrix. It's usually a good starting point. You can obtain the estimates [and associated tests] of the covariance parameters by adding the "/print" subcommand shown below:

Mixed y with time
/fixed time
/print=solution testcov
/random intercept time | subject(id) cov(un).

Best,

Ryan

Gene Maguin wrote

All,
Using Mixed V12, a model such as

Mixed y with time/fixed time/random intercept time | subject(id) cov(ID)

reports a residual variance and an intercept + time variance.

V12+ may be different (and I haven't checked) but I would have expected to
see separate estimates for intercept variance, slope (time) varince and
intercept-slope covariance (as in an LGM). My reading of the documentation
indicates to me that printing G should show the desired estimates but it
doesn't. I either have to be doing something wrong or don't understand how
mixed works. Help please.

Thanks, Gene Maguin

=====================
To manage your subscription to SPSSX-L, send a message to
[e-mail] (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

[Alex Reutter]

The ID covariance matrix is a diagonal matrix with the same variance for each term.  COVTYPE(VC) will give you separate variance estimates for intercept and time.  If you want the off-diagonal term in the G matrix, try COVTYPE(UN).

Alex

From:	Gene Maguin <[hidden email]>
To:	[hidden email]
Date:	05/02/2010 02:56 PM
Subject:	Mixed question
Sent by:	"SPSSX(r) Discussion" <[hidden email]>

---

All,
Using Mixed V12, a model such as

Mixed y with time/fixed time/random intercept time | subject(id) cov(ID)

reports a residual variance and an intercept + time variance.

V12+ may be different (and I haven't checked) but I would have expected to
see separate estimates for intercept variance, slope (time) varince and
intercept-slope covariance (as in an LGM). My reading of the documentation
indicates to me that printing G should show the desired estimates but it
doesn't. I either have to be doing something wrong or don't understand how
mixed works. Help please.

Thanks, Gene Maguin