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 (October 2009)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:   Mon, 12 Oct 2009 09:49:49 +0200
Reply-To:   =?GB2312?B?R2FyY6iqYS1HcmFuZXJv?= <>
Sender:   "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:   =?GB2312?B?R2FyY6iqYS1HcmFuZXJv?= <>
Subject:   Re: T-test Clarification
In-Reply-To:   <COL117-W523335E2090F8AE0A0B0ABD9C80@phx.gbl>
Content-Type:   text/plain; charset=GB2312

Hi Dorraj

DorraJ Oet wrote: > > I need clarification for something basic. > > We know that 2 independent samples T-test has the following assumptions > > 1. The scale follows normal distribution. > 2. The 2 categories in the categorical variable is mutually exclusive. > 3. The homogenity of variance assumption has to be met. > > > My question is as follows > > 1. Assuming that the normality is met, what happen if the > homogenity of variance assumption failed? Do we go to a > non-parametric test say Mann Whitney or do we just use the > variance not assumed T-test Statistics in the T-test table? >

Although not a lot of people knows that, Mann-Whitney's test is very sensitive to differences in spread (heterogeneity of variances). Therefore, it is as bad as a T-test when homogeneity of variances is not met. You must use the "variances not assumed" (Welch test is the name, BTW) T-test. > > 1. I had a professor once commented that he will use the variance > not assumed T-test Statistics and that Statistics is the > calculation when the variance is not pooled but for separated > categories. He even mentioned that going from Parametric to > Non-parametrice approach is not correct. Is he right by > commenting these? >

Yes, he is indeed (see above). "Distribution-free" is a misnomer. Non-parametric tests can have some conditions concerning the distribution of the data. For instance, signed ranks Wilcoxon test needs symmetrical distributions, Mann-Whitney's U and Kruskal-Wallis tests are sensitive to differences in shape and or spread of the groups being compared...

HTH, Prof. Marta Garc¨Şa-Granero

-- For miscellaneous SPSS related statistical stuff, visit:

===================== To manage your subscription to SPSSX-L, send a message to LISTSERV@LISTSERV.UGA.EDU (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD

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