One-way ANOVA
Locked
 
|
Hi SPSS Users, I would like to compare data for 20 states. Each has 24 data points. I am using One-way ANOVA. I did a Levene test for homogeneity of variances. I am getting a sig. value of 1.00. How do I interpreted this as not significant. I also did an F-test for comparing group means. I get a highly significant p-value of 0.000. However when I did the Scheffe test to find out which means are significantly different, the sig. values I am getting range from over 0.8 to 1.00. This is sort of contradicts the result of the F-test. Any help will be greatly appreciated. Thanks, Chris. |
|
|
Could you send the means, sds, and ns for each group. Methinks
there is something rotten in Denmark! Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of Christopher Zindi
|
|
|
In reply to this post by Christopher Zindi
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.. |
|
|
No, the problem is with the number of groups. With 19, and 460
df, the critical value of F is 30.78 because Scheffe’ controls for all
possible comparisons, both pairwise and complex. There are too many comparisons,
about 190 pairwise alone. Even using Bonferroni, which is less conservative
than Scheffe’, the alpha would be .000263 per comparison. Even Tukey’s
fails to detect a difference although some are close. Perhaps a step down test
like regw or the Fisher-Hayter test would find some. With this many means, I
would be tempted to try and group them, perhaps geographically, and do selected
tests. Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston From: Jims More
[mailto:[hidden email]]
Need
a Holiday? Win a $10,000 Holiday of your choice. Enter now.. |
|
|
In reply to this post by Christopher Zindi
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.. |
|
|
That would take advantage of chance. It is best to test hypotheses
that are already not derived from the data. Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston From: Jims More
[mailto:[hidden email]]
Need
a Holiday? Win a $10,000 Holiday of your choice. Enter now.. |
|
|
In reply to this post by Christopher Zindi
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.. |
|
|
It should be done based on knowledge of the population, theory,
or previous research, not on the basis of the data itself. Grouping the data by
looking at it and then doing an ANOVA on those groupings only serves to validate
that the groupings are indeed different but it tells you nothing about why such
differences occur. Grouping the states geographically or financially or on some
other external criterion allows you to see if those groupings differ on the
outcome and if so gives you some information about that external criterion. But
if you group them based on the data itself then you are merely confirming the
process and gaining no insight into the data. Dr. Paul R. Swank, Professor and Director of Research Children's Learning Institute University of Texas Health Science Center-Houston From: Jims More
[mailto:[hidden email]]
Need
a Holiday? Win a $10,000 Holiday of your choice. Enter now.. |
|
|
In reply to this post by Christopher Zindi
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.. |
|
|
Jims
This paper may help
Garry Gealde
Business Analytic Ltd.
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Jims More Sent: 12 June 2009 15:33 To: [hidden email] Subject: Re: One-way ANOVA
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.. __________ Information from ESET NOD32 Antivirus, version of virus signature database 4153 (20090613) __________ The message was checked by ESET NOD32 Antivirus. http://www.eset.com |
| Free forum by Nabble | Edit this page |