Message/Author |
|
|
UG Example 7.9 output how to interpret? In Model results: Categorical Latent Variables Means C#1 -0.085 0.096 -0.889 0.374 What does 0.374 indicate here? The two classes solution is problematic? Tech 7 SAMPLE STATISTICS WEIGHTED BY ESTIMATED CLASS PROBABILITIES FOR CLASS 1 Covariances Y1 Y2 Y3 Y4 ________ ________ ________ ________ Y1 1.111 Y2 0.021 0.966 Y3 -0.034 -0.033 0.892 Y4 -0.072 0.010 0.071 1.113 Does the negative covariance value between Y1 and Y3 signify a “Heywood” case? Or it is of correlation? Thanks in advance. My apology if these simple questions appear somewhere else in the discussion board. |
|
|
Q1: It is the p-value Q2: No. Q3: No; Heywood cases have to do with factor analysis. Q4: It is a covariance. |
|
|
Dear Professor Muthen, Thanks for your clarification. This is a follow-up question. I understand it's the p-value, and the number -0.085 indicates the logit odd of being in C1 compared to C2, which can be converted to obtain the probability. However, I still can't get the meaning of P-value here. Also, about the covariance. Do we need to watch out the negative covariances presented like the above in LPA analysis. Because the journal articles and textbooks talks a lot of statistical adequacy (e.g., no negative variance, Bauer & Curran, 2003) and numerous indices such as BIC, LMR and etc. So I was wondering if there was any special meaning behind a negative covariance. Thanks again for your quick reply. |
|
|
The logit p-value refers to the logit being significantly different from 0 or not. It is typically not an interesting test in this context. Instead you want to focus on the probabilities related to the logits - see end of Chapter 14 of the UG. to learn the basics about mixtures, you can watch our Short Course videos on our website. A negative covariance is like a negative correlation - nothing to be concerned about. |
|
|
Dear Professor Muthen, Thanks for the vivid explanation. I will put more time in learning the basics of mixture modeling from UG and the related videos at your website. |
|
Back to top |