Anova and the Internet

11/17/01

This lab is going to be quite a bit different from the others, and should be fun--well, "fun" may be a bit strong. It is designed to give you experience in running the analysis of variance in SPSS, and to show you some of the stuff that's available on the Internet. In terms of the analysis of variance, be sure that you see how it relates to what I said in class on Tuesday.

First I want you to go to the Web pages that I maintain on zoo. There you will find a worked example that uses the Analysis of Variance, and I want you to read through part of that example yourself. The analyses were done in SAS, but you are going to redo them using SPSS. Go to

http://www.uvm.edu/~dhowell/StatPages/StatHomePage.html

(Be sure that you capitalize exactly as I have.) Click on the link to  Examples, and then find your way to the example on Religious Fundamentalism. I want you to read through this example very quickly, up to, and including, the first SAS example of the Anova. When you come to a paragraph that starts "I chose to run Bonferroni tests ..." you can stop reading.  And you can go very lightly over how the data were generated. Do not print out this document, as it is about 8 pages long. (I'll place a copy in the Xerox room.)

In that example you will see a link to the data. Click on that link, and when the data come up, use the Save command to save them to a file on your hard drive. Note the variable names, as given in the description of the study, so that you can tell SPSS how to name them, and note that one is a string variable. If the data look weird, it will be because you're having problems with carriage returns. Just try to import them anyway, and if that doesn't work, I'll show you want to do. 

Now start SPSS and input the data. Then use SPSS to get the group means and to run an analysis of variance on group differences in Optimism. Print out those results so that you can compare them to the SAS results.

I want you to write these up very quickly. Do not write this as if you were describing the study and its results--as I usually ask you to do. What you write should emphasize how the computer-generated analyses correspond to what I talk about in the book, and what I talked about on Tuesday.

The rest of the lab will be devoted to looking at the Internet, I want people to see what is out there and get some additional experience searching for it. Part of this will be old to some, but new to others.

I want you to see a variety of types of material.

We can't have everyone doing everything at the same time, so do the following in some vaguely random order. Do 4 of the following 5 items and hand in something so that I can see what you did,  or thought, or what you didn't understand.

• Go to http://www.dartmouth.edu/~matc/X10/java/anova/Anova.html. This program shows you what an F distribution looks like when the means are equal, and when they are unequal. In class on 10/25/01 we calculated power with m₁ = 24.00, m₂ = 16.50, for a common standard deviation = 12.03, and n = 100. Set those values up on the display, and draw some samples. (We calculated power to be approximately .70. Does that seem reasonable here?) Move the means around and see what the distributions look like. (I had trouble using this yesterday, but some days it works--don't spend a lot of time trying to get there.
• Go to http://www.dartmouth.edu/~chance/chance_news/current_news/current.html. Read over this page quickly and see if you see anything that could be made into an interesting example.
• Go to the Data and Story Library at Carnegie-Mellon and see what they have to offer. What do you find interesting here?
• Pick a research area that interests you and find at least two useful references. For this you will need one of the major search engines. I would suggest http://www.google.com . Alternatively, you can just click on the Search icon (a magnifying glass) on the Netscape menu and chose one of those search engines.
• Do a quick search for articles on statistical power analysis and the analysis of variance. (That's one topic, not two.)
• One of the multiple comparison procedures that I recommend to students is the Tukey test. Search the web and tell my how you would find out about it.
• I spoke with a graduate student yesterday who needed to know about testing for mediation effects. Find a quick reference.

Last revised: 11/14/01