Warnings concerning a multilevel analysis with dichotomous outcomes

[Elisabeth]

Hello :)

Actually I'm doing my Master thesis and therefor I'm running a multilevel
analysis with clustered data, but actually I get some warnings.

I have 103 participants and each of them answered up to 21 questionnaires (3
times daily). I have some missing data, but I ordered the data in the
vertical format, so I'm loosing around 10% of the measurements.
If I run the null model for a binary outcome (nominal scale; 0,1 format), to
check what proportion of the variance is caused by clustering of the data,
the model is running but I get two warnings, and I'm not sure if I can
ignore them.

I get the following warnings:

glmm: One or more records are not used in the analysis because they have one
or more fields with invalid or missing values.
glmm: Valid values for events (target) and trials variables are non-negative
and positive integers respectively, and the number of trials cannot be less
than the number of events.

If I rerun the model after adding predictors the same warnings are displayed
but the analysis is nevertheless running..


I use the following syntax for my model:


GENLINMIXED
  /DATA_STRUCTURE SUBJECTS=Person REPEATED_MEASURES=MP_timec
COVARIANCE_TYPE=AR1
  /FIELDS TARGET=ND_Fortschritt_GMc_Cutoff TRIALS=NONE OFFSET=NONE
  /TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=LOGIT
  /FIXED USE_INTERCEPT=TRUE
  /RANDOM USE_INTERCEPT=TRUE SUBJECTS=Person
COVARIANCE_TYPE=VARIANCE_COMPONENTS SOLUTION=FALSE
  /BUILD_OPTIONS TARGET_CATEGORY_ORDER=DESCENDING
INPUTS_CATEGORY_ORDER=DESCENDING
    MAX_ITERATIONS=100 CONFIDENCE_LEVEL=95 DF_METHOD=RESIDUAL COVB=ROBUST
PCONVERGE=0.000001(ABSOLUTE)
    SCORING=0 SINGULAR=0.000000000001
  /EMMEANS_OPTIONS SCALE=ORIGINAL PADJUST=LSD.

I would be very grateful if somebody could help me with my question.
Thanks a lot!

Elisabeth






--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

=====================
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]

Sounds like you are using genlinmixed. I haven't used this for a while but what a beast! I think the first message, which is a warning but not an error, is because your DV has missing data. I think that if those cases are excluded that warning will disappear.  I also think the second message is related to missing data as well.

Gene Maguin




-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Elisabeth
Sent: Monday, July 27, 2020 10:06 AM
To: [hidden email]
Subject: Warnings concerning a multilevel analysis with dichotomous outcomes

Hello :)

Actually I'm doing my Master thesis and therefor I'm running a multilevel analysis with clustered data, but actually I get some warnings.

I have 103 participants and each of them answered up to 21 questionnaires (3 times daily). I have some missing data, but I ordered the data in the vertical format, so I'm loosing around 10% of the measurements.
If I run the null model for a binary outcome (nominal scale; 0,1 format), to check what proportion of the variance is caused by clustering of the data, the model is running but I get two warnings, and I'm not sure if I can ignore them.

I get the following warnings:

glmm: One or more records are not used in the analysis because they have one or more fields with invalid or missing values.
glmm: Valid values for events (target) and trials variables are non-negative and positive integers respectively, and the number of trials cannot be less than the number of events.

If I rerun the model after adding predictors the same warnings are displayed but the analysis is nevertheless running..


I use the following syntax for my model:


GENLINMIXED
  /DATA_STRUCTURE SUBJECTS=Person REPEATED_MEASURES=MP_timec
COVARIANCE_TYPE=AR1
  /FIELDS TARGET=ND_Fortschritt_GMc_Cutoff TRIALS=NONE OFFSET=NONE
  /TARGET_OPTIONS DISTRIBUTION=BINOMIAL LINK=LOGIT
  /FIXED USE_INTERCEPT=TRUE
  /RANDOM USE_INTERCEPT=TRUE SUBJECTS=Person COVARIANCE_TYPE=VARIANCE_COMPONENTS SOLUTION=FALSE
  /BUILD_OPTIONS TARGET_CATEGORY_ORDER=DESCENDING INPUTS_CATEGORY_ORDER=DESCENDING
    MAX_ITERATIONS=100 CONFIDENCE_LEVEL=95 DF_METHOD=RESIDUAL COVB=ROBUST
PCONVERGE=0.000001(ABSOLUTE)
    SCORING=0 SINGULAR=0.000000000001
  /EMMEANS_OPTIONS SCALE=ORIGINAL PADJUST=LSD.

I would be very grateful if somebody could help me with my question.
Thanks a lot!

Elisabeth






--
Sent from: http://spssx-discussion.1045642.n5.nabble.com/

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