Multilevel hurdle using NBH option PreviousNext
Mplus Discussion > Categorical Data Modeling >
 S.Arunachalam posted on Friday, July 22, 2016 - 9:16 am
Respected Prof. Muthen

First the new book is wonderful. An eye opener for some models that I never thought could be estimated using Mplus!

Could example in "Table 6.7, p. 273: Hurdle model regression using the NBH" be estimated for multilevel data please? i.e. I have a three level model with two cluster variables.

Is it possible to run this using Mplus? If so any reference would be very useful.

Further kindly comment if there is an advantage in using two-part model using NBT option rather than using NBH option for this case of three level Negative binomial hurdle model.
 Bengt O. Muthen posted on Friday, July 22, 2016 - 2:41 pm
Great that you find our new book useful.

Count outcomes are available for two-level and type=twolevel complex but not yet for three-level.

Regarding your last question I am not sure what the relative advantages would be for 3-level - not sure if there would be different sets of random effects.
 S.Arunachalam posted on Friday, July 22, 2016 - 10:05 pm
Thank you Prof. Muthen.

Gentle Requests:
Would it be possible to include this in V8 please?
Further as Bayes is better for multilevel/hierarchical models, could V8 have Bayes for count models (single level and multilevel) please.
These two additions would be tremendously useful and also help stay within the Mplus platform avoiding using Bugs or STATA for these models, which are very common now due to richer data being available.
 Bengt O. Muthen posted on Saturday, July 23, 2016 - 12:34 pm
We have count for Bayes high on our list but cannot promise when we get to it given competing developments.
 S.Arunachalam posted on Saturday, July 23, 2016 - 7:45 pm
Thank you Prof. Muthen. As always wishing the very best for the new developments. Can't wait for V8!!
 Keelan Meade posted on Tuesday, August 07, 2018 - 6:41 am
Dear Prof. Muthen,

Does MPlus v.8 allow count outcomes for three-level models? If not, could you tell me if there is a plan to include this option in an upcoming revision of MPlus?


*This is my first post and I'm not sure if the MPlus discussion board etiquette is to start a new thread or to post on a relevant older thread. My apologies if I should have started a new thread.
 Tihomir Asparouhov posted on Tuesday, August 07, 2018 - 12:25 pm
No - unless you are able to collapse one of the levels to multivariate model (e.g. if you have only 5 units on level 1 you can use a multivariate two-level model with 5 count variables instead of a univariate three level model), see User's Guide example 9.12.

You can alternatively run the two nesting levels (lever 2 and 3) separately with a two-level model and then retain the level with stronger influence.
 Keelan Meade posted on Wednesday, August 08, 2018 - 2:33 am
Thank you for the advice Tihomir. I think I may be able to run the analysis using the first option.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message