multilevel mixed linear models - error

[Iwona Pil]

Hello,
I have the following problem. I conduct multilevel mixed linear models on 32 subjects clustered in 8 groups with one independent covariate "sequence_length" and one dependent variable "quality", both corrected for normality.

MIXED quality_SQR WITH sequence_length_LN
  /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0,
    ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
  /FIXED=sequence_length_LN | SSTYPE(3)
  /METHOD=REML
  /RANDOM=INTERCEPT sequence_length_LN | SUBJECT(group) COVTYPE(UN).

I get the following message:
Iteration was terminated but convergence has not been achieved. The MIXED procedure continues despite this warning. Subsequent results produced are based on the last iteration. Validity of the model fit is uncertain.

When I exclude the free term for random effect, the model is calculated.
I new to this analysis, so please provide simple explanations...
Please help!
Thank you,

[Ryan]

A few brief comments:

1. The normality assumption has to do with the errors. I would not transform any variables before examining a graphical plot of the residuals.

2. You should know that the variance component will be biased because you only have 8 level-2 units ("group"). You could treat "group" as a fixed effect, but the interpretation will change.

3. Do you believe the linear effect of the covariate ("sequence_length_LN") varies across subjects ("group")? If not, remove the covariate from the RANDOM statement.

Ryan

> On Nov 13, 2013, at 7:56 AM, iwopil <[hidden email]> wrote:
>
> Hello,
> I have the following problem. I conduct multilevel mixed linear models on 32
> subjects clustered in 8 groups with one independent covariate
> "sequence_length" and one dependent variable "quality", both corrected for
> normality.
>
> MIXED quality_SQR WITH sequence_length_LN
>  /CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1)
> SINGULAR(0.000000000001) HCONVERGE(0,
>    ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
>  /FIXED=sequence_length_LN | SSTYPE(3)
>  /METHOD=REML
>  /RANDOM=INTERCEPT sequence_length_LN | SUBJECT(group) COVTYPE(UN).
>
> I get the following message:
> Iteration was terminated but convergence has not been achieved. The MIXED
> procedure continues despite this warning. Subsequent results produced are
> based on the last iteration. Validity of the model fit is uncertain.
>
> When I exclude the free term for random effect, the model is calculated.
> I new to this analysis, so please provide simple explanations...
> Please help!
> Thank you,
>
>
>
> --
> View this message in context: http://spssx-discussion.1045642.n5.nabble.com/multilevel-mixed-linear-models-error-tp5723023.html
> Sent from the SPSSX Discussion mailing list archive at 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