Clustering Effect with SPSS 17 Complex Samples

[Mesfin Mulatu]

Hello All,

I am going to ask a beginner's question, so please bear with me.

I am working with large survey data on characteristics of substance abuse
treatment facilities throughout the united states. Data are collected on all
known facilities that receive government funding. We suspect that there may
be significant state-level clustering effects, given that each state
controls licensing of and disbursement of funding to treatment facilities
within its territories.

We just purchased SPSS 17 with complex samples module, and I am wondering if
someone can help me setup  the complex samples analytical plan. In these
data systems, there are no stratification or weight variables. We have state
codes. Any advice on how to go about it?

Thank you.

Mesfin

=====================
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

[Nichols, David P.]

Hello Mesfin,

If you have data for every known treatment facility in the population of interest, then technically you have no sampling error, so I'll assume that you want to treat these facilities as a sample from an infinite population of potential facilities. In order to create an analysis plan file or to run any of the analytical procedures in the Complex Samples module, you have to have a weight variable. Under these circumstances, I would simply compute one equal to 1 for all cases. Then, in order to create the analysis plan file using the Analysis Preparation Wizard, do the following in the menus:

Select Analyze>Complex Samples>Prepare for Analysis

You'll want the default "Create a plan file" option in the first screen of the wizard. Specify a name and desired location for the file, then click Next.

In the next screen (Stage 1: Design Variables), move your state identifier variable into the Clusters listbox, and your newly created weight variable into the Sample Weight box. Click Next.

In the next screen (Stage1: Estimation Method), uncheck the checkbox for "Use finite population correction (FPC) when estimating variance under simple random sampling assumption" and click Finish.

You've now got your analysis plan file, and whenever you want to analyze those data (or any other data with that structure and the same variable names for cluster IDs and weights) you simply specify that as the Plan file to use in the gating dialog that will come up when you select any of the analysis procedures in the CS module.

This single-stage WR setup is very handy to know, as it is what is most commonly used with published data from U. S. government surveys. Although these typically involve both stratification and clustering variables, and the weight variables are not uniformly 1, the process would differ only in specifying strata in the Stage 1: Design Variables screen, and you might want to leave the default use of the FPC when estimating variance under SRS assumptions in effect if the weights sum to the actual population size (if they're normalized to a mean of 1, as is often the case, you'd probably want to uncheck the box in order to be able to get design effect estimates and have access to all types of test statistics in the analysis procedures).

David Nichols - Master Statistician
SPSS Inc.
Email: [hidden email]
Tel:   [phone] / Fax: [phone]



        From: "Mesfin S. Mulatu" <[hidden email]>
        Date: Thu, 25 Sep 2008 13:33:43 -0500
        Subject: Clustering Effect with SPSS 17 Complex Samples
       
       




        Hello All,
       
        I am going to ask a beginner's question, so please bear with me.
       
        I am working with large survey data on characteristics of substance abuse
        treatment facilities throughout the united states. Data are collected on all
        known facilities that receive government funding. We suspect that there may
        be significant state-level clustering effects, given that each state
        controls licensing of and disbursement of funding to treatment facilities
        within its territories.
       
        We just purchased SPSS 17 with complex samples module, and I am wondering if
        someone can help me setup  the complex samples analytical plan. In these
        data systems, there are no stratification or weight variables. We have state
        codes. Any advice on how to go about it?
       
        Thank you.
       
        Mesfin

====================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