PG851: GLM--ANOVA

This blog is for the use of students enrolled in either section of PG851. It is intended to be a forum for asking, discussing, clarifying, and helping.

Saturday, January 19, 2008

Assumptions of normality - One-way ANOVA hmwk

Katie, Jenel or Dr. Clements
I am working on the assignment for next week and I am having some difficulty in analyzing the data set for violations in the assumption of normality. I have obtained values for the z(skew) = .450 and z(kurtosis) = -.988 but I don't know what to do with them as I don't quite understand what it means to use a p value for this value. First of all, did I need obtain the skew and kurtosis values from the descriptive statistics of this sample set? Secondly, did I need to compute the standardized scores for each of these? Thirdly, would scores, as listed above, indicate violations? I was having a difficult time determining this from the book or the notes.
Thanks
Kris

posted by Kris Thomas @ 5:33 PM  

3 Comments:

  • At 9:33 AM, Blogger Mari said…

    Hi Kris--

    Remember that z scores are sample independent, and thus the critical value for z is the same whether you have a sample size of 20 or 20 thousand.

    In this test, the calculated z scores for skew and for kurtosis are compared to the critical value for z, keeping in mind our one important caveat.

    As described in your notes, you will want to use a p value of .01 or .001 (rather than the customary .05) for tests of skew and kurtosis. This is because these tests are known to be quite conservative. You can get the critical value of z that corresponds to these p values from your notes, or if you prefer, many introductory stats texts include a z distribution table and you can also find such information on the internet (search for z distribution or critical values for z).

    As with any stat, if the calculated value for z does not exceed the cutoff value, you must conclude that you cannot reject the null hypothesis (in this case, that skew and/or kurtosis do not significantly differ from zero).

     

  • At 4:29 PM, Blogger christina and jeff said…

    Where do you find the critical value of z? From a regular z-score table?

     

  • At 8:05 PM, Blogger Mari said…

    You can get the critical value of z that corresponds to these p values from your notes, or if you prefer, many introductory stats texts include a z distribution table and you can also find such information on the internet (search for z distribution or critical values for z).

     

Post a Comment

<< Home