comparing participant and excluded participant characteristics

[karen lewis]

How do I perform a chi-square test with participation and ses characteristics (e.g., education level, poverty status) as my two catagorical variables?  I have a data set that excludes some of the participants from the study because of missing data. I want to compare these two groups across SES variables (e.g., to control for non response bias) but I don’t know where to start.  I am consulting the SPSS survival manual ( Pallant) on how to use chi square, but her examples do not cover this scenario.  I assume that I must start by separating out the participants with missing data from those with a complete set of data for each of the SES variables.  I do not know how to do this (I’ve tried “select cases”, but it did not give me the correct number of missing cases.  It gave me 123, when the number should be 71).  Point and click is preferred, but I will try syntax (I’m a bit rusty on the steps for using syntax).

 

Karen

[Steve Simon, P.Mean Consulting]

Karen Lewis wrote:

> How do I perform a chi-square test with participation and ses
> characteristics (e.g., education level, poverty status) as my two
> catagorical variables?  I have a data set that excludes some of the
> participants from the study because of missing data. I want to compare
> these two groups across SES variables (e.g., to control for non response
> bias) but I don’t know where to start.  I am consulting the SPSS
> survival manual ( Pallant) on how to use chi square, but her examples do
> not cover this scenario.  I assume that I must start by separating out
> the participants with missing data from those with a complete set of
> data for each of the SES variables.  I do not know how to do this (I’ve
> tried “select cases”, but it did not give me the correct number of
> missing cases.  It gave me 123, when the number should be 71).  Point
> and click is preferred, but I will try syntax (I’m a bit rusty on the
> steps for using syntax).

Run crosstabs on your data (ANALYZE | DESCRIPTIVE STATISTICS |
CROSSTABS). Be sure to click on the STATISTICS button on the dialog box
and check the CHISQUARE option.

For the record, though, the chi-square test is inefficient when you have
ordinal data (I assume SES is ordinal). You might want to consider using
the ordinal logistic regression option with SES as the dependent
variable and participation as the factor. Select ANALYZE | REGRESSION |
ORDINAL from the menu.

Steve Simon, Standard Disclaimer
Sign up for The Monthly Mean, the newsletter that
dares to call itself "average" at www.pmean.com/news
"Data entry and data management issues with examples
in IBM SPSS," Tuesday, August 24, 11am-noon CDT.
Free webinar. Details at www.pmean.com/webinars

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

[Bruce Weaver]

Steve Simon, P.Mean Consulting wrote

Karen Lewis wrote:

> How do I perform a chi-square test with participation and ses
> characteristics (e.g., education level, poverty status) as my two
> catagorical variables?  I have a data set that excludes some of the
> participants from the study because of missing data. I want to compare
> these two groups across SES variables (e.g., to control for non response
> bias) but I don’t know where to start.  I am consulting the SPSS
> survival manual ( Pallant) on how to use chi square, but her examples do
> not cover this scenario.  I assume that I must start by separating out
> the participants with missing data from those with a complete set of
> data for each of the SES variables.  I do not know how to do this (I’ve
> tried “select cases”, but it did not give me the correct number of
> missing cases.  It gave me 123, when the number should be 71).  Point
> and click is preferred, but I will try syntax (I’m a bit rusty on the
> steps for using syntax).

Run crosstabs on your data (ANALYZE | DESCRIPTIVE STATISTICS |
CROSSTABS). Be sure to click on the STATISTICS button on the dialog box
and check the CHISQUARE option.

For the record, though, the chi-square test is inefficient when you have
ordinal data (I assume SES is ordinal). You might want to consider using
the ordinal logistic regression option with SES as the dependent
variable and participation as the factor. Select ANALYZE | REGRESSION |
ORDINAL from the menu.

Steve Simon, Standard Disclaimer
Sign up for The Monthly Mean, the newsletter that
dares to call itself "average" at www.pmean.com/news
"Data entry and data management issues with examples
in IBM SPSS," Tuesday, August 24, 11am-noon CDT.
Free webinar. Details at www.pmean.com/webinars

Another (possibly easier) way to take advantage of the ordinal nature of the categories for SES is by using the test of "linear-by-linear" association that you see in the chi-square output from CROSSTABS.  For more info on that test, see Dave Howell's note on it.

   http://www.uvm.edu/~dhowell/methods7/Supplements/OrdinalChiSq.html

The title says "Treatment of Missing Data", but the article is about analysis of data with ordered categories.

[John F Hall]

Karen

 

Steve and Bruce are statistically way above me, but it looks to me as if you're trying to run comparisons of participants against non-participants on SES variables.  What precisely do you mean by missing and exclusion?  Do you have SES data for non-participants?  Do you have a variable for?

 

participant "Yes" "No"  or is this your shorthand for missing cases on the other variables in the survey?

 

If not, you need to create one with an intermediate step.

 

If you send me (an extract of) your SPSS saved file, I'll see if can be of more help, but I much prefer syntax to menus: that way it's simpler, quicker and I can see what I'm doing.  My SPSS tutorials will help to get rid of some of your syntax rust!   Some of what you need is not on the site yet, but I have drafts of tutorials covering SELECT IF etc. which I can send you.  Incidentally, Pallant is an excellent book for desperate dissertation writers, but there are important aspects of surveys and SPSS she doesn't cover (see my reviews of the 2001 and 2005 editions)

 

 

John Hall

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Tuesday, July 27, 2010 9:22 PM

Subject: comparing participant and excluded participant characteristics

How do I perform a chi-square test with participation and ses characteristics (e.g., education level, poverty status) as my two catagorical variables?  I have a data set that excludes some of the participants from the study because of missing data. I want to compare these two groups across SES variables (e.g., to control for non response bias) but I don’t know where to start.  I am consulting the SPSS survival manual ( Pallant) on how to use chi square, but her examples do not cover this scenario.  I assume that I must start by separating out the participants with missing data from those with a complete set of data for each of the SES variables.  I do not know how to do this (I’ve tried “select cases”, but it did not give me the correct number of missing cases.  It gave me 123, when the number should be 71).  Point and click is preferred, but I will try syntax (I’m a bit rusty on the steps for using syntax).

 

Karen

[John F Hall]

Karen

 

Received OK.  Will inspect your file and get back to you.  I've copied this to the list so they can see your reply, but not your data (which is safe with me!). 

 

I have one or two others I've been helping out with problem formulation, complex derived variables, multiple response, file structures etc. (all well off topic for most of the SPSSX list statistical and technical enquiries, but highly relevant to questionnaire surveys) and they are happy for me to use summaries of their problems and my solutions as case studies on the site, anonymously if they prefer.  If we come up with anything of general interest to survey practice, data capture or file design, would you be willing to do the same?

 

This is just like being back at work 20 years ago.

 

John

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Wednesday, July 28, 2010 2:06 PM

Subject: RE: comparing participant and excluded participant characteristics

Hello John,

 

Well, all three of you all are hands down above me J  I have attached my dataset.

 

You’ve described perfectly what I need.  I am trying to run SES characteristic comparisons of participants vs. non-participants.  The non-participants are designated as such because they have missing data files (you’ll see that some of the cells for the various cases are empty). I do not know how to separate out those with missing data from those that have complete data.  BTW, if you need to select variables you can choose any one that you want since you are just demonstrating how I do this.  I am in a rush to get a kid to camp, but wanted to get back to you.  Forgive me if there are typos in this message.  I’m typing on the fly.

 

 

Thanks,

Karen

 

---

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of John F Hall
Sent: Tuesday, July 27, 2010 11:29 PM
To: [hidden email]
Subject: Re: comparing participant and excluded participant characteristics

 

Karen

 

Steve and Bruce are statistically way above me, but it looks to me as if you're trying to run comparisons of participants against non-participants on SES variables.  What precisely do you mean by missing and exclusion?  Do you have SES data for non-participants?  Do you have a variable for?

 

participant "Yes" "No"  or is this your shorthand for missing cases on the other variables in the survey?

 

If not, you need to create one with an intermediate step.

 

If you send me (an extract of) your SPSS saved file, I'll see if can be of more help, but I much prefer syntax to menus: that way it's simpler, quicker and I can see what I'm doing.  My SPSS tutorials will help to get rid of some of your syntax rust!   Some of what you need is not on the site yet, but I have drafts of tutorials covering SELECT IF etc. which I can send you.  Incidentally, Pallant is an excellent book for desperate dissertation writers, but there are important aspects of surveys and SPSS she doesn't cover (see my reviews of the 2001 and 2005 editions)

 

 

John Hall

----- Original Message -----

From: [hidden email]

To: [hidden email]

Sent: Tuesday, July 27, 2010 9:22 PM

Subject: comparing participant and excluded participant characteristics

 

How do I perform a chi-square test with participation and ses characteristics (e.g., education level, poverty status) as my two catagorical variables?  I have a data set that excludes some of the participants from the study because of missing data. I want to compare these two groups across SES variables (e.g., to control for non response bias) but I don’t know where to start.  I am consulting the SPSS survival manual ( Pallant) on how to use chi square, but her examples do not cover this scenario.  I assume that I must start by separating out the participants with missing data from those with a complete set of data for each of the SES variables.  I do not know how to do this (I’ve tried “select cases”, but it did not give me the correct number of missing cases.  It gave me 123, when the number should be 71).  Point and click is preferred, but I will try syntax (I’m a bit rusty on the steps for using syntax).

 

Karen