Date: Fri, 12 Jun 2009 07:32:37 -0700
Reply-To: Jims More <morejims@yahoo.com.au>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Jims More <morejims@yahoo.com.au>
Subject: Re: One-way ANOVA
Content-Type: multipart/alternative;
Now I am enlighten, Paul.
Thank you so much for your explanation.
Regards,
Jims
--- On Fri, 12/6/09, Swank, Paul R <Paul.R.Swank@uth.tmc.edu> wrote:
From: Swank, Paul R <Paul.R.Swank@uth.tmc.edu>
Subject: RE: One-way ANOVA
To: "Jims More" <morejims@yahoo.com.au>, "SPSSX-L@LISTSERV.UGA.EDU" <SPSSX-L@LISTSERV.UGA.EDU>
Received: Friday, 12 June, 2009, 2:26 PM
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:morejims@yahoo.com.au]
Sent: Friday, June 12, 2009 9:15 AM
To: SPSSX-L@LISTSERV.UGA.EDU; Swank, Paul R
Subject: Re: One-way ANOVA
Paul wrote:
>>>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.
Jims Replied:
>>>Can he not do cluster analysis first of the 20 states?
Paul replied:
<<<That would take advantage of chance. It is best to test hypotheses that are already not derived from the data.
I think grouping geographycally is subjective, right? Do you prefer subjective grouping?
Jims
--- On Fri, 12/6/09, Swank, Paul R <Paul.R.Swank@uth.tmc.edu> wrote:
From: Swank, Paul R <Paul.R.Swank@uth.tmc.edu>
Subject: Re: One-way ANOVA
To: SPSSX-L@LISTSERV.UGA.EDU
Received: Friday, 12 June, 2009, 1:28 PM
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:morejims@yahoo.com.au]
Sent: Friday, June 12, 2009 7:38 AM
To: SPSSX-L@LISTSERV.UGA.EDU; Swank, Paul R
Subject: Re: One-way ANOVA
Can he not do cluster analysis first of the 20 states?
--- On Thu, 11/6/09, Swank, Paul R <Paul.R.Swank@uth.tmc.edu> wrote:
From: Swank, Paul R <Paul.R.Swank@uth.tmc.edu>
Subject: Re: One-way ANOVA
To: SPSSX-L@LISTSERV.UGA.EDU
Received: Thursday, 11 June, 2009, 2:44 PM
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:morejims@yahoo.com.au]
Sent: Thursday, June 11, 2009 5:20 AM
To: SPSSX-L@LISTSERV.UGA.EDU; Swank, Paul R
Subject: Re: One-way ANOVA
Is it not a sample size problem?
--- On Wed, 10/6/09, Swank, Paul R <Paul.R.Swank@uth.tmc.edu> wrote:
From: Swank, Paul R <Paul.R.Swank@uth.tmc.edu>
Subject: Re: One-way ANOVA
To: SPSSX-L@LISTSERV.UGA.EDU
Received: Wednesday, 10 June, 2009, 2:08 PM
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:SPSSX-L@LISTSERV.UGA.EDU] On Behalf Of Christopher Zindi
Sent: Tuesday, June 09, 2009 4:08 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: One-way ANOVA
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.
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now..
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now..
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now..
Need a Holiday? Win a $10,000 Holiday of your choice. Enter now.http://us.lrd.yahoo.com/_ylc=X3oDMTJxN2x2ZmNpBF9zAzIwMjM2MTY2MTMEdG1fZG1lY2gDVGV4dCBMaW5rBHRtX2xuawNVMTEwMzk3NwR0bV9uZXQDWWFob28hBHRtX3BvcwN0YWdsaW5lBHRtX3BwdHkDYXVueg--/SIG=14600t3ni/**http%3A//au.rd.yahoo.com/mail/tagline/creativeholidays/*http%3A//au.docs.yahoo.com/homepageset/%3Fp1=other%26p2=au%26p3=mailtagline
[text/html]
