Cases in use when specifying a multivariate response model

Welcome to the forum for MLwiN users. Feel free to post your question about MLwiN software here. The Centre for Multilevel Modelling take no responsibility for the accuracy of these posts, we are unable to monitor them closely. Do go ahead and post your question and thank you in advance if you find the time to post any answers!

Remember to check out our extensive software FAQs which may answer your question: http://www.bristol.ac.uk/cmm/software/s ... port-faqs/
Post Reply
hannakim
Posts: 2
Joined: Sun May 14, 2017 1:00 pm

Cases in use when specifying a multivariate response model

Post by hannakim » Sun May 14, 2017 1:46 pm

Dear forum,

I would like to ask if anyone of you knows what "(3428 of 3810 cases in use)" and "UNITS: school_long: 73 (of 73) in use / student_long: 1905 (of 1905) in use" mean.

I am referring to page 222 from the MLwiN3.0 manual, where it shows the equations window using the worksheet gcsemv1.ws supplied with the MLwiN software(I attach a screenshot that I made by imitating the example with MLwin 2.36). The model consists of two response variables(written and csework) and each has 202 and 180 missing values.

Now I am curious since there is only one line that shows the number of cases used. How should I take this? I mean, 3428 cases are only those that have no missing on the two dependent variables, but in the same time, the 'UNITS' info tells me that no person has been removed.

Does this mean that no person has been deleted automatically and the parameters have been calculated using casewise deletion while the loglikelihood or covariances are calculated using only those 3428 cases? Or does it show that the cases with missing dependent variables are not included in the analysis?

If that is the case, is there any other way to incorporate the cases with missing DV's in a bivariate multilevel model? I am looking for a way to specify such a model, and hope this could be possible with MLwiN. I will be looking forward to your response.
Thank you in advance :)

With best regards,
Hanna
Attachments
mlwin.png
mlwin.png (25.05 KiB) Viewed 111 times

ChrisCharlton
Posts: 890
Joined: Mon Oct 19, 2009 10:34 am

Re: Cases in use when specifying a multivariate response model

Post by ChrisCharlton » Mon May 15, 2017 12:38 pm

"(3428 of 3810 cases in use)" means that of the 3810 rows in the data 3428 are used in the model. Any rows where there are missing values in the variables used in the model are excluded. Only the row containing missing is excluded, so a student can contribute information on coursework if they have missing written data and vice versa.

"student_long: 1905 (of 1905)" means that of the 1905 student units in the data none are entirely missing (i.e. all students have a score for written or coursework or both). If the first number was less than 1905 then it would mean that there are students for which there is no data being used in the model.

Similarly "school_long: 73 (of 73)" means that of the 73 schools defined in the data there are none where all the students have been excluded.

If you were to instead run the models with MCMC in MLwiN then the missing values in the responses will be filled in (see chapter 18 of the MCMC guide - http://www.bristol.ac.uk/cmm/software/m ... nuals.html).

I hope this makes sense.

hannakim
Posts: 2
Joined: Sun May 14, 2017 1:00 pm

Re: Cases in use when specifying a multivariate response model

Post by hannakim » Mon May 15, 2017 4:17 pm

Dear Chris,
this is really helpful! I am so relieved to know that no data missing only one DV is deleted listwise.

Just for caution's sake, is there any difference in estimating multivariate multilevel models between MLwiN 2.36 and 3.00? I see that the UNIT info does not show up in MLwiN 2.36 when I run models by myself, but I wonder if it would be just an additional feature of the ver. 3.00 and have no influence on the results.

ChrisCharlton
Posts: 890
Joined: Mon Oct 19, 2009 10:34 am

Re: Cases in use when specifying a multivariate response model

Post by ChrisCharlton » Mon May 15, 2017 4:23 pm

There is no difference in estimation between version 2.36 and 3.00. The unit information is available via Model->Hierarchy Viewer in both versions, however it was also added to the equation window in 3.00 to improve visibility of the information.

Post Reply