Re: where can I learn about frequencies adjusting for population characteristics
Locked
Re: where can I learn about frequencies adjusting for population characteristics
 
|
Gene One way might involve devising some complex weighting, about which I know next to nothing. Another way would be to produce conditional frequency tables (or contingency tables) and barcharts for sub-populations from each group. If you have the raw data files, Excel spreadsheets or SPSS saved files for your groups and are prepared to send them to me (in confidence) I can have a look at your data and run some analyses for you. As everyone knows, a picture (or chart) is worth a thousand words, and, as my old boss Mark Abrams used to say, “If it’s worth saying, you can say it in percentages.” I’m copying the SPSS list into this as there are some really clever guys out there who may have helpful suggestions. John F Hall (Mr) [Retired academic survey researcher] Email: [hidden email] Website: www.surveyresearch.weebly.com Start page: www.surveyresearch.weebly.com/spss-without-tears.html From: Survey Research Methods Section of the ASA [mailto:[hidden email]] On Behalf Of Gene Shackman Hi all
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To subscribe/unsubscribe SRMSNet: http://listserv.umd.edu/cgi-bin/wa?A0=srmsnet&D=0&F=&H=0&O=T&S=&T=1 SRMS website: http://www.amstat.org/sections/srms/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
Re: where can I learn about frequencies adjusting for population characteristics
|
The SPSSINC RAKE extension command adjusts
sample weights to match specified control totals in up to ten dimensions.
That command and the Python Essentials prerequisite can be downloaded
from the SPSS Community website (www.ibm.com/developerworks/spssdevcentral).
Maybe you would want to use the average of these totals across the
two datasets or take one as definitive.
Jon Peck (no "h") aka Kim Senior Software Engineer, IBM [hidden email] phone: 720-342-5621 From: John F Hall <[hidden email]> To: [hidden email], Date: 04/12/2013 09:59 AM Subject: Re: [SPSSX-L] where can I learn about frequencies adjusting for population characteristics Sent by: "SPSSX(r) Discussion" <[hidden email]> Gene
One way might involve devising some complex weighting, about which I know next to nothing.
Another way would be to produce conditional frequency tables (or contingency tables) and barcharts for sub-populations from each group.
If you have the raw data files, Excel spreadsheets or SPSS saved files for your groups and are prepared to send them to me (in confidence) I can have a look at your data and run some analyses for you.
As everyone knows, a picture (or chart) is worth a thousand words, and, as my old boss Mark Abrams used to say, “If it’s worth saying, you can say it in percentages.”
I’m copying the SPSS list into this as there are some really clever guys out there who may have helpful suggestions.
John F Hall (Mr) [Retired academic survey researcher]
Email: johnfhall@... Website: www.surveyresearch.weebly.com Start page: www.surveyresearch.weebly.com/spss-without-tears.html
From: Survey Research Methods Section
of the ASA [mailto:SRMSNET@...]
On Behalf Of Gene Shackman
Hi all
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ To subscribe/unsubscribe SRMSNet: http://listserv.umd.edu/cgi-bin/wa?A0=srmsnet&D=0&F=&H=0&O=T&S=&T=1 SRMS website: http://www.amstat.org/sections/srms/ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
Re: where can I learn about frequencies adjusting for population characteristics
|
I too, like Jon, think that multivariate weighting would be helpful:
equalize qroups by their background characteristics before
computing/tabulating percents or means. Note however that by doing
the re-weighting you move towards quasi-experimantal study away from
natural-observational survey.
Wighting macro "Weighting groups", similar alternative to SPSSINC RAKE extension command, can be found at rivita.ru/spssmacros_en.shtml 12.04.2013 20:21, Jon K Peck пишет:
The SPSSINC RAKE extension command adjusts sample weights to match specified control totals in up to ten dimensions. � That command and the Python Essentials prerequisite can be downloaded from the SPSS Community website (www.ibm.com/developerworks/spssdevcentral). � Maybe you would want to use the average of these totals across the two datasets or take one as definitive. |
Re: where can I learn about frequencies adjusting for population characteristics
|
|
In reply to this post by John F Hall
There are two usual problems with weighting, whenever the weighting
has much impact. First, the tests are no longer completely valid. Second, somebody else might have preferred a different weighting. - Epidemiologists sometimes handle the second by using a national population profile. The important thing is the resulting narrative. Do the simple main effects, using no weights, convey what is going on? If not, then you probably want to display what John suggests -- the frequencies or barcharts for the relevant sub-populations. -- Rich Ulrich Date: Fri, 12 Apr 2013 17:50:06 +0200 From: [hidden email] Subject: Re: where can I learn about frequencies adjusting for population characteristics To: [hidden email] Gene
One way might involve devising some complex weighting, about which I know next to nothing.
Another way would be to produce conditional frequency tables (or contingency tables) and barcharts for sub-populations from each group.
If you have the raw data files, Excel spreadsheets or SPSS saved files for your groups and are prepared to send them to me (in confidence) I can have a look at your data and run some analyses for you.
As everyone knows, a picture (or chart) is worth a thousand words, and, as my old boss Mark Abrams used to say, “If it’s worth saying, you can say it in percentages.”
I’m copying the SPSS list into this as there are some really clever guys out there who may have helpful suggestions.
John F Hall (Mr) [Retired academic survey researcher]
Email: [hidden email] Website: www.surveyresearch.weebly.com Start page: www.surveyresearch.weebly.com/spss-without-tears.html
From: Survey Research Methods Section of the ASA [mailto:[hidden email]] On Behalf Of Gene Shackman Hi all We have some survey data from two different groups. Their population characteristics are slightly different (e.g., age, race/ethnicity, etc.). Where can I learn how to adjust them so that I can present their responses to survey questions, adjusted for their population characteristics? So for example I want to present tables showing what percent from groups 1 and 2 said they were satisfied with ABCDE, after making them comparable for the various population characteristics. That way, people can see, from the table, where there are differences. I know I can do regressions to adjust, and get statistically significant differences, but I want to have easy to read tables too, that non statisticians can read. Where on the web shows how to do this? ... |
|
|
In reply to this post by John F Hall
I cannot find the original post in my inbox, so I will just respond to this post. This post is not intended to be a response to Jon's comments. I am not entirely certain what you are asking...
If you want to adjust for group differences on key characteristics, one simple approach would be to employ a binary logistic regression where group membership is the dependent variable, and the key characteristics for which you want to adjust are the predictors. Output the predicted probabilities and use those as a covariate in the model.
Important point: While you want to adjust for group differences, you should hope to not find significant group differences on those characteristics. That is, when you employ the logistic regression, you should hope that the key characteristics do not significantly/substantially predict group membership. The more they do, the less valid this endeavour becomes. More complex and perhaps optimal approaches have been developed, but I do not have time to expound upon them.
HTH, Ryan On Fri, Apr 12, 2013 at 11:50 AM, John F Hall <[hidden email]> wrote:
|
| Free forum by Nabble | Edit this page |