Mixed Model Quandry - Can anybody help?

[David Allsop]

Dear All

I am trying to run a mixed model for repeated measures with post hoc pairwise comparisons - using the following code:


DATASET ACTIVATE DataSet1.
MIXED  LEA BY Timepoint Cann_Control Study_Code
  /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0,
    ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
  /FIXED=Timepoint Cann_Control Timepoint*Cann_Control | SSTYPE(3)
  /METHOD=REML
  /RANDOM=INTERCEPT Study_Code | SUBJECT(Study_Code) COVTYPE(VC)
  /REPEATED=Timepoint | SUBJECT(Study_Code) COVTYPE(AR1)
  /EMMEANS=TABLES(Cann_Control*Timepoint) COMPARE (Cann_Control) ADJ (BONFERRONI).

The weird thing is, the analysis runs, but the Type III omnibus statistics table is empty in the output - yet the post hoc pairwise comparisons table is populated with values...... I have NO IDEA what could cause this behaviour - so I'm going to post in the raw data for the variables in question below - and hope that somebody can help me troubleshoot this.

Here is the data (study code 1st data point = 1, Cann_Control = 1, Timepoint = 1, LEA = 0.00000000000011900 - just to help you align):

Study_Code Cann_Control Timepoint LEA
1 1 1 0.000000000011900
1 1 2 0.000000000013000
1 1 3 0.000000000011900
1 1 4 0.000000000011200
2 1 1 0.000000000008550
2 1 2 0.000000000009280
2 1 3 0.000000000007380
2 1 4 0.000000000008330
3 1 1 0.000000000012100
3 1 2 0.000000000012000
3 1 3 0.000000000009510
3 1 4 0.000000000012600
4 1 1 0.000000000007390
4 1 2 0.000000000007930
4 1 3 0.000000000007300
4 1 4 0.000000000007520
5 0 1 0.000000000015246
5 0 2 0.000000000010789
5 0 3 0.000000000011582
5 0 4 0.000000000008112
6 1 1 0.000000000011000
6 1 2 0.000000000012000
6 1 3 0.000000000014300
6 1 4 0.000000000015200
7 0 1 0.000000000008411
7 0 2 0.000000000006268
7 0 3 0.000000000007079
7 0 4
8 1 1 0.000000000006900
8 1 2 0.000000000005100
8 1 3 0.000000000008910
8 1 4 0.000000000009800
9 0 1 0.000000000007354
9 0 2 0.000000000006473
9 0 3 0.000000000004069
9 0 4 0.000000000004469
10 0 1 0.000000000006573
10 0 2 0.000000000006105
10 0 3 0.000000000006937
10 0 4 0.000000000005991
11 0 1 0.000000000005822
11 0 2 0.000000000007360
11 0 3 0.000000000008113
11 0 4 0.000000000009862
12 0 1 0.000000000006451
12 0 2 0.000000000007761
12 0 3 0.000000000009403
12 0 4 0.000000000007942
13 0 1
13 0 2
13 0 3
13 0 4
14 1 1 0.000000000010400
14 1 2 0.000000000012100
14 1 3 0.000000000016300
14 1 4 0.000000000015200
15 1 1 0.000000000017100
15 1 2 0.000000000014200
15 1 3 0.000000000020700
15 1 4 0.000000000015500
16 1 1 0.000000000014100
16 1 2 0.000000000014100
16 1 3 0.000000000012500
16 1 4 0.000000000013200
17 1 1 0.000000000007580
17 1 2 0.000000000008030
17 1 3 0.000000000009200
17 1 4 0.000000000010700
19 0 1 0.000000000007267
19 0 2 0.000000000008529
19 0 3 0.000000000007603
19 0 4 0.000000000009126
21 0 1 0.000000000009077
21 0 2 0.000000000010857
21 0 3 0.000000000010654
21 0 4 0.000000000013857
23 0 1 0.000000000005468
23 0 2 0.000000000005032
23 0 3 0.000000000008667
23 0 4 0.000000000008744
24 0 1 0.000000000007980
24 0 2 0.000000000008061
24 0 3 0.000000000014414
24 0 4 0.000000000009455

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

[Maguin, Eugene]

I may be misunderstanding your analysis but I'm very surprised 1) to see that study_code, which appears to be the id variable being named on the IV list and 2) to see that there is a random statement used. As I understand your experimental design structure, there were two groups of subects (Cann_Control) and all subjects were assessed by LEA at each of four time points (TimePoint). Subjects weren't assessed by LEA multiple times, e.g., in multiple trials, at each time point were they? If not, then it seems that you have an ordinary one-between (Cann-Control with two levels) and one-within (TimePoint with four levels). You could do this in GLM repeated measures, in Mixed as a growth curve or in mixed as a repeated measures. Perhaps there is something I don't understand but this is how I understand it at this point.
Gene Maguin

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of David Allsop
Sent: Monday, March 20, 2017 12:58 AM
To: [hidden email]
Subject: Mixed Model Quandry - Can anybody help?

Dear All

I am trying to run a mixed model for repeated measures with post hoc pairwise comparisons - using the following code:


DATASET ACTIVATE DataSet1.
MIXED  LEA BY Timepoint Cann_Control Study_Code
  /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0,
    ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
  /FIXED=Timepoint Cann_Control Timepoint*Cann_Control | SSTYPE(3)
  /METHOD=REML
  /RANDOM=INTERCEPT Study_Code | SUBJECT(Study_Code) COVTYPE(VC)
  /REPEATED=Timepoint | SUBJECT(Study_Code) COVTYPE(AR1)
  /EMMEANS=TABLES(Cann_Control*Timepoint) COMPARE (Cann_Control) ADJ (BONFERRONI).

The weird thing is, the analysis runs, but the Type III omnibus statistics table is empty in the output - yet the post hoc pairwise comparisons table is populated with values...... I have NO IDEA what could cause this behaviour - so I'm going to post in the raw data for the variables in question below - and hope that somebody can help me troubleshoot this.

Here is the data (study code 1st data point = 1, Cann_Control = 1, Timepoint = 1, LEA = 0.00000000000011900 - just to help you align):

Study_Code Cann_Control Timepoint LEA
1 1 1 0.000000000011900
1 1 2 0.000000000013000
1 1 3 0.000000000011900
1 1 4 0.000000000011200
2 1 1 0.000000000008550
2 1 2 0.000000000009280
2 1 3 0.000000000007380
2 1 4 0.000000000008330
3 1 1 0.000000000012100
3 1 2 0.000000000012000
3 1 3 0.000000000009510
3 1 4 0.000000000012600
4 1 1 0.000000000007390
4 1 2 0.000000000007930
4 1 3 0.000000000007300
4 1 4 0.000000000007520
5 0 1 0.000000000015246
5 0 2 0.000000000010789
5 0 3 0.000000000011582
5 0 4 0.000000000008112
6 1 1 0.000000000011000
6 1 2 0.000000000012000
6 1 3 0.000000000014300
6 1 4 0.000000000015200
7 0 1 0.000000000008411
7 0 2 0.000000000006268
7 0 3 0.000000000007079
7 0 4
8 1 1 0.000000000006900
8 1 2 0.000000000005100
8 1 3 0.000000000008910
8 1 4 0.000000000009800
9 0 1 0.000000000007354
9 0 2 0.000000000006473
9 0 3 0.000000000004069
9 0 4 0.000000000004469
10 0 1 0.000000000006573
10 0 2 0.000000000006105
10 0 3 0.000000000006937
10 0 4 0.000000000005991
11 0 1 0.000000000005822
11 0 2 0.000000000007360
11 0 3 0.000000000008113
11 0 4 0.000000000009862
12 0 1 0.000000000006451
12 0 2 0.000000000007761
12 0 3 0.000000000009403
12 0 4 0.000000000007942
13 0 1
13 0 2
13 0 3
13 0 4
14 1 1 0.000000000010400
14 1 2 0.000000000012100
14 1 3 0.000000000016300
14 1 4 0.000000000015200
15 1 1 0.000000000017100
15 1 2 0.000000000014200
15 1 3 0.000000000020700
15 1 4 0.000000000015500
16 1 1 0.000000000014100
16 1 2 0.000000000014100
16 1 3 0.000000000012500
16 1 4 0.000000000013200
17 1 1 0.000000000007580
17 1 2 0.000000000008030
17 1 3 0.000000000009200
17 1 4 0.000000000010700
19 0 1 0.000000000007267
19 0 2 0.000000000008529
19 0 3 0.000000000007603
19 0 4 0.000000000009126
21 0 1 0.000000000009077
21 0 2 0.000000000010857
21 0 3 0.000000000010654
21 0 4 0.000000000013857
23 0 1 0.000000000005468
23 0 2 0.000000000005032
23 0 3 0.000000000008667
23 0 4 0.000000000008744
24 0 1 0.000000000007980
24 0 2 0.000000000008061
24 0 3 0.000000000014414
24 0 4 0.000000000009455

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