Documentation of multiple imputation in SPSS 17
Locked
 
|
Hi,
I am looking for a documentation of multiple imputation in SPSS 17 - and I can only find articles that deal with Schafer's NORM. But as far as I understood there are several possibilities of implementing MI. Unfortunately I understand only half of the FCS-description in the documentation of the SPSS-algorithms. Additionally this documentation is rather short and doesn't help me to realize where SPSS differs from NORM and where not. Can anyone refer me to more detailed descriptions of how SPSS implents MI? I would appreciate every reference! Thank you in advance -- Katrin |
Re: Documentation of multiple imputation in SPSS 17
|
|
Hello Katrin,
The method used by SPSS differs from NORM in the sense that NORM specifies one multivariate (normal) model for the data, while SPSS specifies a conditional model for each variable separately. When all variables are continuous, the methods used by SPSS and by NORM are said to be identical, although I've heard from colleagues that there is some discussion about that, but I have not actually seen this discussion in the literature. Two articles that discuss the method used by SPSS are:
I hope this will help you out.
Best regards,
Joost van Ginkel Van: SPSSX(r) Discussion namens Zerpy Verzonden: wo 15-7-2009 17:21 Aan: [hidden email] Onderwerp: Documentation of multiple imputation in SPSS 17 Hi, |
are my results real or an artifact of my method
|
|
In reply to this post by Zerpy
Listees,
I am looking at the effect of an intervention in the emergency department of hospitals that allows the doctor to look up patient information online. I want to see if there is a change in patient utilization of services, such as emergency department, their pcp, outpatient services, inpatient services, depending on whether their doctor accessed the online system when they visited the emergency room. Hopefully people use all services less when their doctor accesses this system. Here's the issue. The data just look too perfect, like an upside down letter 'v', with the sides a little curved. Here's my method. Is this result due to my method or is it an accurate representation of the data. My method: I'm setting the point at which the doctor accessed the online information as zero. I'm calculating the time between the doctor accessing the system (this is CREATED.1, a date, in the syntax) and the date of service for that medical services (fromdatesvc_s). I'm calculating the number of months between these two dates, rounding to the nearest month. I've used the Date and Time Wizard to derive this formula. Is this formula resulting in a mathematical artifact that is giving me such perfect results? Is there something I'm missing? COMPUTE entry_diff=RND((CREATED.1 - fromdatesvc_s) / (30.4375 * time.days(1))). Also, on a related note, how does the Date and Time Wizard come up with the divisor here (30.4375*time.days(1))? As always, any help is much appreciated. Thanks Matt Matthew Pirritano, Ph.D. Research Analyst IV Medical Services Initiative (MSI) Orange County Health Care Agency (714) 568-5648 ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
| Free forum by Nabble | Edit this page |