Levene's test
This was Marisol's question posted to the regression blog by mistake...
I have three questions:1) Should we report the F and p value under the Levene's test in the write up for t-test2) When we are checking for the assumption of homogenity of variance, do we have to calculate F max or we can say that Levene's test shown that there is equal variances. Is Levene's test related to the assumption of homogenity of variance? If we have to calculate F max, from where in my printout do I get the S squared (or the necessary information to calculate it)?3) In class it was mentioned that it is worthless to report degrees of freedom as (1, 145) for t-test and that the appropiate way is to say just t(145), does that apply to one-wayANOVA when you are reporting the F?
Posted by Marisol De Jesus at 3:45 PM
I have three questions:1) Should we report the F and p value under the Levene's test in the write up for t-test2) When we are checking for the assumption of homogenity of variance, do we have to calculate F max or we can say that Levene's test shown that there is equal variances. Is Levene's test related to the assumption of homogenity of variance? If we have to calculate F max, from where in my printout do I get the S squared (or the necessary information to calculate it)?3) In class it was mentioned that it is worthless to report degrees of freedom as (1, 145) for t-test and that the appropiate way is to say just t(145), does that apply to one-wayANOVA when you are reporting the F?
Posted by Marisol De Jesus at 3:45 PM
posted by Bettyvs @ 9:17 PM
2 Comments:
At 10:47 PM,
Unknown said…
For question 1:
The F and p values for Levene's are not necessary to report because you are only using them to see if you use equal variances assumed or not assumed.
For question 2:
I do not believe you need to calculate F max unless cannot assume equal variance. Yes, Levene's test is related to the assumption of homogeneity of variance. I think that all of the various formulas you would need to calculate F max are in your handouts. It might be on a different page, but I am pretty sure all the ones you need are in there.
For question 3:
(I'm just using your numbers as examples.) For a t-test it would be t(145) and for one-way ANOVA, it would be F(1, 145). In other words, include both types of degrees of freedom for F.
At 12:26 PM,
Mari said…
Please see slide 14 of the factorial ANOVA handout posted on portico if you need a refresher about how to get SPSS to calculate variance for you. Note that you will have to split your file by the groups being analyzed in order to get the variance per cell.
Note that you do not have to follow up the Levene's test with an Fmax under the following conditions: (a) Levene's is not significant, or (b) Levene's is significant, and you use post-hoc comparisons that account for this. Otherwise, you will want to compute Fmax, and the formula for that is provided both in your handouts and on p. 234 of your text. As you may have guessed by where these items are located, it was not necessary to compute Fmax to receive full credit on the oneway assignment: it was fine to stick with Levene's test.
You must ALWAYS report both degrees of freedom for an F test. A t-test only has one number for its degree of freedom, and you should only report the single number. An ANOVA will always have two, and you must always report them both.
Post a Comment
<< Home