| Message/Author |
|
|
|
|
I am getting an error [shown at the end of the message]. In another sample, I get the same error except that one ends with ".....I on Suphistress."
Looking for solutions on how to modify syntax below.
DEFINE: Center age_1 (GRANDMEAN);
Center schyr (GRANDMEAN);
Center isel_sum (GRANDMEAN);
Center zcssmean (GRANDMEAN);
lowstress = zcssmean - .284;
histress = zcssmean + .284;
iselxcss = zcssmean*isel_sum;
suplowstress = isel_sum*lowstress;
suphistress = isel_sum*histress;
MODEL:
i s | lcrp1_10@0 lcrp2_10@1 lcrp3_10@2;
i on suplowstress suphistress isel_sum zcssmean age_1 race2 schyr;
s on suplowstress suphistress isel_sum zcssmean age_1 race2 schyr;
OUTPUT: STDYX sampstat tech3 tech4 MODINDICES (3.84);
Warnings in output :
WARNING: THE SAMPLE COVARIANCE OF THE INDEPENDENT VARIABLES IS SINGULAR. PROBLEM INVOLVING VARIABLE SUPHISTR.
WARNING: THE SAMPLE COVARIANCE IS SINGULAR.
WARNING: THE SAMPLE CORRELATION OF HISTRESS AND LOWSTRESS IS 1.000.
THE MODEL ESTIMATION TERMINATED NORMALLY.
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THE MODEL MAY NOT BE IDENTIFIED. CHECK YOUR MODEL. PROBLEM INVOLVING THE FOLLOWING PARAMETER:
Parameter 25, S ON SUPHISTRESS |
|
|
|
|
Your variable definition
lowstress = zcssmean - .284;
histress = zcssmean + .284;
causes a correlation of 1 between the 2 variables. This causes the Warning message about the sample covariance/correlation matrix. You can't have both of these variables in the model. |
|
|
|
|
Thank you!!
I implemented your suggestion (deleted lowstress and suplowstress variables and ran model with only histress and suphistress variables, and vice versa). Getting readable output but model fit is very poor and indices are identical in both models (RMSEA=.956, CFI=.042, TLI=-.474, SRMR=.174 in histress and SRMR=.170 in lowstress model). Regression coefficiencts and p-values are suspiciously identical for both interaction terms, suphistress and suplowstress (b= .018, p= .017).
I am also getting the following warnings:
WARNING IN MODEL COMMAND: Variable is uncorrelated with all other variables: LOWSTRESS (same for histress)
WARNING IN MODEL COMMAND: Variable is uncorrelated with all other variables : ISELxCSS (same for histress model)
WARNING IN MODEL COMMAND: At least one variable is uncorrelated with all other variables in the model. Check that this is what is intended (in both models)
WARNING: THE SAMPLE COVARIANCE IN SINGULAR (in both models)
WARNING: THE SAMPLE CORRELATION OF ZCSSMEAN AND LOWSTRESS IS 1.000 (same for histress model)
Questions:
1. Did I adequately implement your suggestion?
2. How can I improve model fit?
3. Is this the right way to examine simple slope? I am trying to examine the effect of stress on CRP change (a protein) over 3 time points based on high and low levels of social support. |
|
|
|
|
I should mention that "Zcssmean" is centered at the mean (as shown in the original syntax in the first message) - I notice in the univariate sample statistics that the mean of zcssmean is 0.00 and the SD is .284 and mean of histress is .284 and SD is .284, which sounds right. It should be estimating high levels of stress at 1 SD above the mean. But I wonder if somehow this is causing the warning pertaining to the sample correlation being 1.00 between zcssmean and histress, and that led to my question #3 above.
Same issue applies to lowstress as that's estimated as 1 SD below the mean. |
|
|
|
|
It sounds like you still have sample correlations of 1 in your data. You have to first focus on not getting correlations of 1 in your sample correlation matrix - such sample correlations are not allowed in the analysis. Do a Type=Basic analysis so you can check this.
If 2 variables differ only by a constant they will have correlation=1. |
|
|
| Back to top |
