PY603 - Advanced Statistics II

2007-12-06: Projects will be due on Wednesday, December 12. If you need an extension let me know.
2007-12-06: Exam 2 will be posted on the website on Friday, and should be turned in by Monday. It will cover material from Chapter 4 to Chapter 9. The exam will be closed-book, so you should not look at your notes, homeworks, prior exams, or anything related to this class between the time you first open up the exam to the time you email it to py603@bama.ua.edu. If you are inadvertantly exposed to information related to the class during this time you must contact me so that I can give you new questions that are untainted by whatever information you have heard.
2007-12-06: Made a correction to the HW8 key. I had misinterpreted the coefficients on my effect codes. Thanks Abbie for catching that!
2007-11-29: We'll be talking about some wise comments of Jacob Cohen in Stat Lunch tomorrow. Come by 204 GP if you get a chance.
2007-11-15: Jamie, Mark Klinger, and Nancy Rhodes will be talking about analyzing reaction time data tomorrow at Stat-Lunch. Come by 172 GP if you'd like to hear more!
2007-11-09: I've updated the regression notes to include new sections on how to test the significance of specific simple slope coefficients. The new content is located on pages 54-57.
2007-10-26: Avani will have limited availability between Novermber 9 and November 21. She will not be holding the Stat Chat on Tuesdays on November 13 and November 20, and may take a little more time to respond to your questions.
2007-09-28: Brittany is taking over my job! She actually searched and found the new link to Wooley's suppression article, which is posted below. Guess it's time for me to retire to Jamaica and carve wind chimes for a living.
2007-09-06: Jamie will be giving a presentation at Contemporary Issues this week (Friday 9/7 at 1pm in 204 GP) entitled "Contemplations on p-values." Please come if you have an interest!
2007-09-04: Avani's office hours have been officially changed to Tuesday from 6-7pm. Feel free to contact her at other times as well if that time doesn't work for you.
2007-09-04: I've added a new link up above to saved chat logs. Whenever I have a Yahoo messenger chat with someone I will ask them if it would be ok to post it on the website. When they do, I'll upload the file into the chat directory so that anyone can take a look at it. We will also save copies of Avani's office hour sessions to that directory.
2007-08-30: Below are some of the formulas relating to p-values that you can find in Excel, along with explanations of how they work.

TDIST(t statistic, df, tails): This formula gives you the p-value for a t statistic and degrees of freedom. The t statistic can be either positive or negative.

TINV(probability, degrees of freedom): This formula will give you the value from the t distribution where the two-tailed p-value is equal to "probability." It always gives you a positive number. If we label the result of the formula to be X, the probability is the proportion of the t distribution that is less than -X plus the proprotion of the distribution that is greater than X. If you want to know the critical value when generating a confidence interval with confidence level equal to Y, you would use the formula TINV(1-Y, df). Therefore, if you wanted the critical value for a .95 confidence interval that had 39 df, you would use the formula TINV(.05, 39).
NORMSDIST(Z statistic): This formula gives you the proportion of the Z distribution that is to the left of the Z statistic. Notice that this is only the p-value when the Z statistic is negative and you are interested in a 1-tailed test. If your Z statistic is positive, then you have to either put -Z into the formula, or else subtract the probability you get from 1 to get your p-value. For example, to get the p-value for a Z of 3 you would need to use either the formula NORMSDIST(-3) or the formula 1 - NORMSDIST(3). If you want the two-tailed p-value, you would need to multiply this result by two. Therefore, if you wanted the two-tailed p-value for a Z statistic of 2.5, you would use the formula 2*NORMSDIST(-2.5).
NORMSINV(probability): This formula gives you value from the Z distribution that is greater than the given probability. If the probability is less than .5, then the resulting Z will be negative. If the probability is greater than .5, then the resulting Z will be positive. If you want to know the critical value when generating a confidence interval with confidence level equal to Y, you would need to use the formula NORMSINV(1-(1-Y)/2). Therefore, if you wanted the critical value for a .95 confidence interval, you would use the formula NORMSINV(1-(1-.95)/2), which would be equal to NORMSINV(.975).

2007-08-27: The notes on ANOVA and Excel that were posted prior to today were actually older versions. I have updated the files on the website.
2007-08-27: Avani has changed the chat hours to be Tuesday at 5pm because of student conflicts.
2007-08-23: Where does the yahoo pager ID come from? An old latin prayer for enlightenment:

O sibili si ergo
Fortibuses in ero
O nobili demis trux
Si wat senem
Cowzen dux

Links

To WebCT login

http://webct.ua.edu/

Related to data and data sets

http://lib.stat.cmu.edu/datasets/ - Statlib data set archive

Related to multiple regression

http://www.graphpad.com/articles/Multicollinearity.htm - Single-page explanation of multicollinearity by Harvey Motulsky
http://ericae.net/ft/tamu/supres.htm - Kristin K. Woolley's article on suppressor variables.

Related to testing model assumptions

http://www.stat.umn.edu/arc/ - Website for the free software package ARC

Related to mediation

http://davidakenny.net/cm/mediate.htm - David A. Kenny's mediation website
http://www.public.asu.edu/~davidpm/ripl/mediate.htm - David P. MacKinnon's mediation website
http://www.unc.edu/~preacher/sobel/sobel.htm - Online Sobel test calculator

To other web resources

http://www.adobe.com/products/acrobat/alternate.html - Adobe acrobat (for reading notes)
http://members.aol.com/johnp71/javastat.html - A very nice index to interactive statistical web pages
http://www.randomizer.org/ - A web page that generates lists of random numbers that you can use when designing studies

http://www.stat-help.com - Free statistical consulting over the internet (provided by Jamie, et al.)

http://www.allexperts.com - Free help with anything else!