Message/Author |
|
|
Hi! We are trying to examine the covariation of 2 variables using growth curve modeling. However, the variables have a ceiling effect. When we look at a median split analysis, there is a definite pattern of the trajectories covarying, so we are hoping to capture this pattern using the full sample as well. We read that there were ways to deal with such data; specifically we read in a couple of places about Tobit, but were not able to find code to incorporate into our LCM analyses. We'd appreciate any help with this, whether it is using a Tobit or another method to deal with the ceiling effect. Thanks! Adar |
|
|
Tobit modeling is obtained using the CENSORED option of the VARIABLE command. You may also want to consider two-part modeling. See Papers on the website where there is a section on two-part modeling. |
|
|
Thank-you Linda, We were able to run the analyses, but have a couple of questions: 1) Did we specify the model correctly - we constrained the slope to be linear i1 s1 | mapp1@0 mapp2@1 mapp3@2 i2 s2 | int1@0 int2@1 int3@2 This seems to constrain the slope and we are not sure if the censored command works with the constrained slopes as the point is to be able to re-assess the slope with the censored command fixing the spread of these variables. 2)How do we interpret the output? What should I be looking for in terms of the descriptives for these curves? etc. thanks, Adar |
|
|
1. I do not understand your question. 2. The plots are for the model estimated means of the observed censored variable. |
|
|
Hi Linda, We were able to run a model, however, we are not sure how to deal with a warning that emerged. We are not sure if this warning is important to resolve or whether we can report the relations between the intercepts and slopes and disregard the warning. The warning is right above where it says that the model terminated normally Thanks you so much for your help, Adar CENSORED are s1mapps (a) s2mapps (a)s3mapps (a) s1pis (a) s2pis (a) s3pis (a); Analysis: convergence = .001; type = missing ; ALGORITHM=INTEGRATION; integration=5; start= 500 20; MODEL: i1 s1 | S1mapps@0 S2mapps@1 S3mapps@2; i2 s2 | S1pis@0 S2pis@1 S3pis@2; Unperturbed starting value run did not converge. 18 perturbed starting value run(s) did not converge. 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 IDENTIFICATION. THE CONDITION NUMBER 0.696D-21. PROBLEM INVOLVING PARAMETER 20. THE MODEL ESTIMATION TERMINATED NORMALLY |
|
|
This message should not be ignored. Please send the full otupt and your license number to support@statmodel.com. |
|
Back to top |