Easy question
Locked
 
|
This is probably a really easy question, but forgive me I am new to SPSS.
I'm trying to examine survey data for differences between this year and last year. Basically, I want to see if the percentage of people responding "yes" to a question this year is different than last year. Is crosstabs the best way to test this? Also, what if teh data has more than two categories? If a the crosstabs test is significant, how do I determine where the real differences are? Are there post-hoc tests available for crosstabs? Thanks for you help. |
|
|
The statistics coming with CROSSTABS are probably the most adequate
response to your question, especially CHI SQUARE and the measures based in it, like PHI. There are several measures available, but their use depends on the specific question you want to answer. For instance, whether there is a significant difference between the two years, or whether you can predict one year once you know the other one, and so on. I recommend reading the SPSS User's Guide on this. Hector -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Amanda Sent: 08 July 2007 00:15 To: [hidden email] Subject: Easy question This is probably a really easy question, but forgive me I am new to SPSS. I'm trying to examine survey data for differences between this year and last year. Basically, I want to see if the percentage of people responding "yes" to a question this year is different than last year. Is crosstabs the best way to test this? Also, what if teh data has more than two categories? If a the crosstabs test is significant, how do I determine where the real differences are? Are there post-hoc tests available for crosstabs? Thanks for you help. |
|
|
I don't think there is a way to do any 'post hoc' tests in chi-square
analysis. One has to just 'eye' the cells to conclude where the significant result might be coming from. Bozena Bozena Zdaniuk, Ph.D. University of Pittsburgh UCSUR, 6th Fl. 121 University Place Pittsburgh, PA 15260 Ph.: 412-624-5736 Fax: 412-624-4810 email: [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Hector Maletta Sent: Sunday, July 08, 2007 5:49 PM To: [hidden email] Subject: Re: Easy question The statistics coming with CROSSTABS are probably the most adequate response to your question, especially CHI SQUARE and the measures based in it, like PHI. There are several measures available, but their use depends on the specific question you want to answer. For instance, whether there is a significant difference between the two years, or whether you can predict one year once you know the other one, and so on. I recommend reading the SPSS User's Guide on this. Hector -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Amanda Sent: 08 July 2007 00:15 To: [hidden email] Subject: Easy question This is probably a really easy question, but forgive me I am new to SPSS. I'm trying to examine survey data for differences between this year and last year. Basically, I want to see if the percentage of people responding "yes" to a question this year is different than last year. Is crosstabs the best way to test this? Also, what if teh data has more than two categories? If a the crosstabs test is significant, how do I determine where the real differences are? Are there post-hoc tests available for crosstabs? Thanks for you help. |
|
|
The easiest way is to display the residuals which are the difference between
expected and actual counts. -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Zdaniuk, Bozena Sent: Sunday, July 08, 2007 4:40 PM To: [hidden email] Subject: Re: Easy question I don't think there is a way to do any 'post hoc' tests in chi-square analysis. One has to just 'eye' the cells to conclude where the significant result might be coming from. Bozena Bozena Zdaniuk, Ph.D. University of Pittsburgh UCSUR, 6th Fl. 121 University Place Pittsburgh, PA 15260 Ph.: 412-624-5736 Fax: 412-624-4810 email: [hidden email] -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Hector Maletta Sent: Sunday, July 08, 2007 5:49 PM To: [hidden email] Subject: Re: Easy question The statistics coming with CROSSTABS are probably the most adequate response to your question, especially CHI SQUARE and the measures based in it, like PHI. There are several measures available, but their use depends on the specific question you want to answer. For instance, whether there is a significant difference between the two years, or whether you can predict one year once you know the other one, and so on. I recommend reading the SPSS User's Guide on this. Hector -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Amanda Sent: 08 July 2007 00:15 To: [hidden email] Subject: Easy question This is probably a really easy question, but forgive me I am new to SPSS. I'm trying to examine survey data for differences between this year and last year. Basically, I want to see if the percentage of people responding "yes" to a question this year is different than last year. Is crosstabs the best way to test this? Also, what if teh data has more than two categories? If a the crosstabs test is significant, how do I determine where the real differences are? Are there post-hoc tests available for crosstabs? Thanks for you help. |
|
|
In reply to this post by Amanda-20
The overall chi-square can signal that proportions
differ but does not pinpoint wherein the differences lie. You can partition chi-square following rules for partitioning. This is discussed in Agresti's texts. CHAID implements a form of testing that collapses categories or keeps them distinct. SPSS Ctables has some built-in significance testing. -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Amanda Sent: Saturday, July 07, 2007 10:15 PM To: [hidden email] Subject: Easy question This is probably a really easy question, but forgive me I am new to SPSS. I'm trying to examine survey data for differences between this year and last year. Basically, I want to see if the percentage of people responding "yes" to a question this year is different than last year. Is crosstabs the best way to test this? Also, what if teh data has more than two categories? If a the crosstabs test is significant, how do I determine where the real differences are? Are there post-hoc tests available for crosstabs? Thanks for you help. |
|
In reply to this post by ViAnn Beadle
At 06:39 PM 7/8/2007, Zdaniuk, Bozena wrote:
>I don't think there is a way to do any 'post hoc' tests in chi-square >analysis. One has to just 'eye' the cells to conclude where the >significant result might be coming from. At 08:17 PM 7/8/2007, ViAnn Beadle wrote: >The easiest way is to display the residuals which are the difference >between expected and actual counts. Right. Most useful of all, display the adjusted standardized residual. (It's option ASRESID on CELLS subcommand, if you're writing syntax; or, available from the Cells dialog in the menus.) Look for cells with values above +2 or below -2. Frequently, row or column percents are helpful; which is more helpful depends on how you're thinking of your data. (You can have both, of course, but counts plus one of the percentages plus ASRESID is about as much as you can do without making the cells look cluttered.) That lets you spot cells that are notably large, or notably small, within their row (or column). Percentages are not redundant with ASRESID. The percentages measure an effect size; the ASRESID is an estimate of the statistical significance of that effect. Don't put too much weight on a high or low percentage value unless ASRESID 'says' it's credible; it may be random variation in a very small cell. On the other hand, in a cell with a lot of members, you may get a strong ASRESID value when the cell isn't enough higher or lower than expected, to have much practical importance. |
|
|
All,
Could I just be having a Monday morning thinking clearly problem with respect to Amanda's posting, the result of going to a great music festival? As I read her question, she has a dichotomous variable and wants to compare the proportion of yeses this year with the proportion from last year. If her variable were something like height, everybody would say, "Use a paired sample t-test". Nobody would say "Use correlations". Crosstabs measures association between polychotomous variables. I think she needs a test of paired proportions. Gene Maguin |
|
|
In reply to this post by Amanda-20
This is just a z-test of proportions, is it not?
Mark ----- Original Message ----- From: Gene Maguin [[hidden email]] Sent: 07/09/2007 09:56 AM To: [hidden email] Subject: Re: Easy question All, Could I just be having a Monday morning thinking clearly problem with respect to Amanda's posting, the result of going to a great music festival? As I read her question, she has a dichotomous variable and wants to compare the proportion of yeses this year with the proportion from last year. If her variable were something like height, everybody would say, "Use a paired sample t-test". Nobody would say "Use correlations". Crosstabs measures association between polychotomous variables. I think she needs a test of paired proportions. Gene Maguin |
|
|
In reply to this post by Amanda-20
If they are truly paired and not two independent samples taken at two points in time, then she can use the McNemar test (non-parametric test for 2 related samples).
Paula -----Original Message----- From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Gene Maguin Sent: Monday, July 09, 2007 9:56 AM To: [hidden email] Subject: Re: Easy question All, Could I just be having a Monday morning thinking clearly problem with respect to Amanda's posting, the result of going to a great music festival? As I read her question, she has a dichotomous variable and wants to compare the proportion of yeses this year with the proportion from last year. If her variable were something like height, everybody would say, "Use a paired sample t-test". Nobody would say "Use correlations". Crosstabs measures association between polychotomous variables. I think she needs a test of paired proportions. Gene Maguin |
|
|
In reply to this post by Mark A Davenport MADAVENP
Hi Gene,
This sounds like a case for the z test of column proportions, supported by SPSS in the tables module. Structure your table such that year defines the columns and your binary variable defines the rows, and then specify column proportions as the summary. In the Statistics tab of the Tables module, you will have the option to check the box for Z tests of Column proportions, the results for which will tell you whether the difference in the proportion of cases answering yes in year 1 is statistically significant relative to the proportion of cases answering yes in year 2. The Z test makes pair-wise comparisons of column proportions within each row of the table. So, assuming you have enough data, this test can be used when the table employs more than 2 columns and 2 rows. HT, John Norton SPSS Inc. ----- Original Message ----- From: Gene Maguin [[hidden email]] Sent: 07/09/2007 09:56 AM To: [hidden email] Subject: Re: Easy question All, Could I just be having a Monday morning thinking clearly problem with respect to Amanda's posting, the result of going to a great music festival? As I read her question, she has a dichotomous variable and wants to compare the proportion of yeses this year with the proportion from last year. If her variable were something like height, everybody would say, "Use a paired sample t-test". Nobody would say "Use correlations". Crosstabs measures association between polychotomous variables. I think she needs a test of paired proportions. Gene Maguin |
Re: FW: Re: Easy question
|
|
In reply to this post by Cloutier, Paula
I just checked my copy of Fleiss, Levin, and Paik (2003) Statistical Methods for Rates and Proportions and chapter 15 covers correlated binary data. Much of it looks to be related to the intraclass correlation. Mark *************************************************************************************************************************************************************** Mark A. Davenport Ph.D. Senior Research Analyst Office of Institutional Research The University of North Carolina at Greensboro 336.256.0395 [hidden email] 'An approximate answer to the right question is worth a good deal more than an exact answer to an approximate question.' --a paraphrase of J. W. Tukey (1962) To: [hidden email] |
Query: Warning "Input Error when reading a case"
|
|
In reply to this post by Cloutier, Paula
Hi, can anyone please help me with this warning message from SPSS. It
appears as if the dataset I am using is corrupt but I can't work out why? *****Warning message********** Warnings Command name: FREQUENCIES Input error when reading a case. This command not executed. ************************************* Thanks very much Jamie Burnett Statistics Manager Ipsos MORI T +44 20 7347 3338 F +44 20 7347 3803 E [hidden email] W www.ipsos-mori.com 79-81 Borough Road, London, SE1 1FY ============================ This e-mail and all attachments it may contain is confidential and intended solely for the use of the individual to whom it is addressed. Any views or opinions presented are solely those of the author and do not necessarily represent those of Ipsos MORI and its associated companies. If you are not the intended recipient, be advised that you have received this e-mail in error and that any use, dissemination, printing, forwarding or copying of this e-mail is strictly prohibited. Please contact the sender if you have received this e-mail in error. Market & Opinion Research International Ltd , Registered in England and Wales No. 948470 , 79-81 Borough Road , London SE1 1FY, United Kingdom, Email: [hidden email] ============================ |
Re: Query: Warning "Input Error when reading a case"
|
|
At 07:54 AM 9/4/2007, Jamie Burnett wrote:
>Hi, can anyone please help me with this warning message from SPSS. It >appears as if the dataset I am using is corrupt but I can't work out >why? > >*****Warning message********** > >Warnings >Command name: FREQUENCIES >Input error when reading a case. >This command not executed. >************************************* This is the second such report that's been posted recently; see also thread "input error when reading a case", Sat, 18 Aug 2007 <10:27:54 -0700> ff. So, conceivably, there's a problem of some kind. In the earlier thread, the problem was with a file created and saved with SPSS 14. Syntax and data were available to re-create the file, and that was a successful solution. Your best bet may be to do the same. I posted some suggestions ("Re: input error when reading a case", Sat, 18 Aug 2007 >14:26:25 -0400>). But as I wrote then, at best they've only a chance of diagnosing or fixing the problem; and, in the earlier case, they didn't. What version of SPSS are you running, under what operating system? If it's 14 or 15, have you put in the latest patches? Have you seen anything similar before? |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |