Impossible EM values from MVA

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

bobpbody
Posts: 2
Joined: Sun Jan 29, 2012 12:36 am

Impossible EM values from MVA

Postby bobpbody » Sun Jan 29, 2012 1:50 am

I used EM on all the variables from a survey that had possible responses from 1-6. How can it have calculated replacement values that are negative or way out of range? There are several values over 10 now. How can that have happened? I used EM with Normal distribution on the missing value analysis. I have SPSS 19.0.0 for Mac.
Thanks for your help
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Impossible EM values from MVA

Postby GerineL » Wed Feb 01, 2012 1:00 pm

what does EM stand for?
and are you sure you don't have errors in entry of the data?
bobpbody
Posts: 2
Joined: Sun Jan 29, 2012 12:36 am

Re: Impossible EM values from MVA

Postby bobpbody » Wed Feb 01, 2012 3:20 pm

EM is expectation-maximization, a method for replacing missing values. I don't have errors in my data, but it appears that the problematic items occur when several items in a row were skipped. Is it possible that the algorithm for replacing values goes haywire when it has to operate on several items in a case that are missing?
Thanks.
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: Impossible EM values from MVA

Postby GerineL » Thu Feb 02, 2012 10:16 am

so with my missing values menu the third window is to give it constrains, have you not done that?
So than you can say that they have to be between 1 and 6, and there cannot be any decimals.

Who is online

Users browsing this forum: No registered users and 1 guest

cron