Message/Author |
|
Lynn posted on Wednesday, February 17, 2010 - 2:05 am
|
|
|
Dear Dr. Muthen, I am using multilevel regression analysis, the model running normally except a warning. Can i fix it and can this results be used? Below is warning message that I had WARNING: THE MLR STANDARD ERRORS COULD NOT BE COMPUTED. THE MLF STANDARD ERRORS WERE COMPUTED INSTEAD. THE MLR CONDITION NUMBER IS -0.433D-04. PROBLEM INVOLVING PARAMETER 24 (PSI, S1). THIS MAY BE DUE TO NEAR SINGULARITY OF THE RANDOM EFFECT VARIANCE/COVARIANCE OR INCOMPLETE CONVERGENCE. The syntax are: VARIABLE: NAMES ARE ...... USEVARIABLES ARE y1 y2 x1 x2 m cluster; WITHIN = x1 x2; BETWEEN = m; CLUSTER IS cluster; CENTERING = GRANDMEAN (x1 x2); ANALYSIS: TYPE = TWOLEVEL RANDOM; MODEL: %WITHIN% s1 | y1 ON x1; s2 | y1 ON x2; s3 | y2 ON x1; s4 | y2 ON x2; %BETWEEN% s1-s4 y1 y2 ON m; OUTPUT: TECH1 TECH8 ; Thank you very much. Kitkat |
|
|
If you find standard errors in your results, they can be trusted. The problem is most likely due to the variances of the random slopes being close to zero or correlations of one between some of the random slopes. Check the results for zero variances and TECH4 for correlations of one. |
|
Mihyun Park posted on Wednesday, May 04, 2011 - 9:10 am
|
|
|
Dear Dr. Linda I had following warning message. I can't understand well about that. Can you tell me more about that? THE MODEL ESTIMATION TERMINATED NORMALLY WARNING: THE LATENT VARIABLE COVARIANCE MATRIX (PSI) IN GROUP MALE IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/ RESIDUAL VARIANCE FOR A LATENT VARIABLE, A CORRELATION GREATER OR EQUAL TO ONE BETWEEN TWO LATENT VARIABLES, OR A LINEAR DEPENDENCY AMONG MORE THAN TWO LATENT VARIABLES. CHECK THE TECH4 OUTPUT FOR MORE INFORMATION. PROBLEM INVOLVING VARIABLE F2. |
|
|
f2 has either a negative variance, a correlation greater or equal to one with another variable, or is linearly related to 2 or more variables. Check TECH4. |
|
|
Hello, I am running the following model: WITHIN = zgphtt zlet1 gender zage momed; BETWEEN = zmeansr; MODEL: %within% s1 | zlet4 on Zgphtt; s2 | zlet4 on gender; zlet4 on zlet1 momed zage; zlet1 with momed gender zage Zgphtt; momed with gender zage Zgphtt; gender with zage Zgphtt; zage with Zgphtt; zlet1 momed gender zage Zgphtt zlet4; [zlet1 momed gender zage Zgphtt]; %between% zLet4 s1 s2 ON zmeansr; zmeansr zlet4 s1 s2; I received the following warning: THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS -0.195D-17. PROBLEM INVOLVING THE FOLLOWING PARAMETER: Parameter 24, %WITHIN%: MOMED Because this is just a level 1 only 'control' variable, I dropped it and reran, I got the same warning with a different variable (zgphtt). At this point I reran with momed back in, but commented out ALL means, variances and covariances; this makes the error go away. I was wondering if the error is associated with the fact that one of my main predictors (gender) is binary and can be ignored? |
|
|
You are bringing your observed exogenous variables into the model by mentioning their means, variances, and covariances. The message is caused because the mean and variance of a binary variable are not orthogonal. You can ignore the message because this is the cause. By the way, you must include all or none of the observed exogenous variables. You cannot include a subset. |
|
|
Thank you! Could you please clarify for me: "you must include all or none of the observed exogenous variables. You cannot include a subset" -are you referring to the fact that I dropped momed to evaluate what would happen, or am I missing some coding? Thanks again! |
|
|
Yes, I was referring to that. If you have 8 covariates, you cannot include only 2 or 3. It must be all 8 or none. If you include a subset, the covariances among those included and those excluded is zero which is not what you would want. |
|
|
Thanks! |
|
Back to top |