LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (January 2003)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Fri, 31 Jan 2003 11:40:56 +0100
Reply-To:     Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Subject:      Re: Reporting 1-tailed t-test results
Content-Type: text/plain; charset=us-ascii

Hi Janet:

> I want to report the results of a 1-tailed independent samples t-test. It > seems that SPSS produces only 2-tailed results. I believe that I can simply > double the 2-tailed significance level, and that is easy enough.

Just the opposite. You need to halve the 2-tailed value if the data go in the right direction (I mean, if they go the same way your hypothesis does). If the data go the other way (queer, but sometimes it happens), then the one tailed p value you are looking for is 1-(2tailed p)/2, and the result will be non significant even if the 2-tailed value was.

One more comment: sometimes, using 1-tailed p values is considered a way of cheating: if you have ended up with a 2 -tailed p-value of 0.06 (p>0.05, non significant, although the topic is far more complicated than this simple statement), the corresponding one-tailed p value is 0.03 (p<0.05, and everybody is happy).

> However, I am not clear about how, if at all, the other data are affected. I > am > specifically concerned about: > > t > Sig.(2-tailed) --> The only one affected > Mean difference > Std. Error Difference > Levene's Test F > Levene's Test Significance >

The rest remains unchanged, with the exception of the confidence interval for the mean difference, which you don't mention (and the one I consider most important).

> If only the significance level for the t-test needs to be changed, I'm all > set. But if changes are required for anything else, is there a > straightforward way to calculate them? If there is not, how do others > report 1-tailed test results?

I strongly recommend you not to use one-tailed p-values. One excelent discussion on the topic can be found in Ware JH, Mosteller F et al. P values. In: Medical uses of statistics. 2nd Ed. New England Journal of Medicine Books.

I also recommend you the following article by Trisha Greenhalgh: How to read a paper: Statistics for the non-statistician. I: Different types of data need different statistical tests.

It's available at: http://bmj.com/cgi/content/full/315/7104/364.

Regards

Marta Garcia-Granero


Back to: Top of message | Previous page | Main SPSSX-L page