Type=Complex and singleton PSU's PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Peter Cabrera-Nguyen posted on Saturday, April 03, 2010 - 5:42 am
Drs. Múthen,

I am analyzing the Collaborative Psychiatric Epidemiology Surveys public use data set, and their website explicitly states to use the clustering variable (SECLUSTR) as the PSU. Unfortunately, SECLUSTR has a maximum value of 2, and therefore there are far too many singleton PSU's to handle them by simply deleting those observations from an analysis.

In Stata 10/SE, one may handle this problem by using the "singleunit" option following the svyset command. According to ICPSR's CPES documentation, the singleunit option should be specified as "singleunit(certainty)". Does Mplus have a similar option for handling singleton PSU's for "type=complex"?. If so, what would it be? If not, how would you advise handling a complex survey dataset in mplus that has a substantial number of singleton PSU's that cannot be deleted?

Thank you in advance for you assistance.

Sincerely,
Peter
 Linda K. Muthen posted on Saturday, April 03, 2010 - 9:35 am
Mplus does not delete singleton PSU's.
 Sara Suzuki posted on Thursday, January 17, 2019 - 10:09 am
How does Mplus treat singleton PSUs?
 Tihomir Asparouhov posted on Thursday, January 17, 2019 - 2:05 pm
You should see this message in the output which answers the question.

WARNING: THE VARIANCE CONTRIBUTION FROM A STRATUM WITH A SINGLE CLUSTER
(PSU) IS BASED ON THE DIFFERENCE BETWEEN THE SINGLE CLUSTER VALUE AND
THE OVERALL CLUSTER MEAN.

In terms of formula (6) in
http://www.statmodel2.com/download/webnotes/mplusnote72.pdf
nh/(nh-1) is replaced by 1 for that single PSU stratum and vh is replaced by the average of vh across all the strata.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: