|
|
Measurement errors for most likely class |
|
Message/Author |
|
|
Hello, I am trying to do 3-step estimation of a distal in an LTA, using Webnote 15 (Version 8, August 5, 2014). My indicators are continuous, i.e. cross-sectionally I have LPAs. For step 3 (as illustrated in Appendix N, p. 30) I need the measurement errors of the most likely class variables. Webnote 15, on page 15, instructs us to fix these parameters at the step 2 log ratios. I am not sure how to get the necessary values for this. Can somebody point me in the right direction? Thank you. |
|
|
Look for this part of the output: Logits for the Classification Probabilities for the Most Likely Latent Class Membership (Column) by Latent Class (Row) 1 2 1 3.533 0.000 2 -3.853 0.000 |
|
|
Hello, I am conducting a simulation study using the manual 3-step ML method. Is there a way to save the "Logits for the Classification Probabilities for the Most Likely Latent Class Membership (Column) by Latent Class (Row)" from the output? Similar to how one would save class probablities using save=cprob. Alternatively, is there is straight forward way to compute these values using the information that is saved in the cprob option? I know that averaging the posterior probabilities by class will produce the "Average Latent Class Probabilities for Most Likely Latent Class Membership (Row) by Latent Class (Column)." However, it seems that the logit estimates are based on the "Classification Probabilities for the Most Likely Latent Class Membership (Column) by Latent Class (Row)." Thank you! |
|
|
There is no convinient way to save them but I think you can use https://www.tandfonline.com/doi/abs/10.1080/10705511.2017.1402334?journalCode=hsem20 to do that. Alternatively, if you can get the "Average Latent Class Probabilities..." from cprob you can do it there. Instead of getting the average you would need to get the "Total Latent Class Probabilities..." from cprob, i.e., don't divide by the number of observations classified in the class. From there you transpose the matrix and standardize the rows so that the values in each row add up to 1. This gives you the table "Classification Probabilities for the Most Likely Latent Class Membership (Column) by Latent Class (Row)." From there you get the logits. A third way is to use https://www.statmodel.com/utility/extractor.shtml to pick the values up directly from the output. You can download it here http://www.statmodel.com/examples/webnotes/web10.zip |
|
Back to top |
|
|