Re: Mixed procedure in partly correlated data

Posted by Rich Ulrich on Nov 20, 2018; 7:33pm
URL: http://spssx-discussion.165.s1.nabble.com/Mixed-procedure-in-partly-correlated-data-tp5737006p5737009.html

Or, Why did those 29 or 33 happen to get measured at least twice?

I would want to compare the "multi-measured" to the rest, to find out if they differ.

If they differ, I would not like of the MIXED model; it buries information.

I would use the multiple measures to take a look at reliability/stability over time (and

age effect, within person); the N of 29 (or 33) is large enough to be useful.

I'm never excited by the complexity of the MIXED when so little information is added,

so I would certainly carry out (if only for my own satisfaction) analyses that used the

first record for each child.

--

Rich Ulrich

---

From: SPSSX(r) Discussion <[hidden email]> on behalf of Maguin, Eugene <[hidden email]>
Sent: Tuesday, November 20, 2018 1:42 PM
To: [hidden email]
Subject: Re: Mixed procedure in partly correlated data

 

I think an important initial question is how/why did so many kids appear only once.
Gene Maguin

-----Original Message-----
From: SPSSX(r) Discussion <[hidden email]> On Behalf Of Erkki Komulainen
Sent: Tuesday, November 20, 2018 11:25 AM
To: [hidden email]
Subject: Mixed procedure in partly correlated data

Hi listers!

I have a data of children tested approximately at the ages of 4, 5 and 6 yrs. N=192. 122 of them appear in the data-matrix only once, 29 appear twice and 4 even three times. Observations are thus partly correlated. The traditional (say) univariate anova goes from the assumption of independent rows in the data. I did the analysis using the mixed procedure where subject-variable is carrying the information of being the same child in different occasion. Most f=122 have unique value in this variable. Any thoughts of using the traditional vs mixed procedure?

Syntax:

MIXED W3YtS BY PIQgrp
  /METHOD=REML
  /PRINT=SOLUTION TESTCOV DESCRIPTIVES
  /FIXED= PIQgrp | SSTYPE(3)
  /EMMEANS = TABLES (PIQgrp)  COMPARE(PIQgrp) REFCAT(1)
   ADJ(LSD )
  /RANDOM = INTERCEPT | Subject(Subject) covtype(ID).

Erkki Komulainen

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

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

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