fi=memo0324

TO: Students in CJ 405
FROM: R. B. Taylor
DATE: 3/24/99
RE: Comments on simple regression homeworks


After reading the last set of homeworks, I have some comments that you might want to think about when preparing future homeworks that describe regression problems, or related analyses.
DESCRIBE THE SAMPLE AND THE VARIABLES. Tell the reader about your data, just as you would if writing a journal article. You need not go on for pages and pages but describe: the data source, how obtained, some features of the sample. This is important to give the reader a sense of the information used, especially when you are describing the individual level data file.
DESCRIBE THE DEPENDENT VARIABLE IN SPECIFIC TERMS. Help the reader make sense of the dependent variable. If it is an index, what goes into it, in specific terms (describe component variables). What is the variation on the outcome (describe min, max, median, mean, sd, for example)? Most importantly, be sure the reader knows what a HIGH score on the outcome means and what a LOW score means.
DESCRIBE B WEIGHT IN TERMS SPECIFIC TO THE PROBLEM. It is not enough to say that "as X goes up one unit, Y goes up .1234 units." Describe the b weight in terms that are RELEVANT to the predictor. If you predictor is age, "going up one unit" means "for each additional year in age." If you have a dummy variable say "as we switch from those not scared of guns (coded 0), to those scared of guns (coded 1), the average score on the outcome ..."
DECIDING STRONG VS. WEAK B WEIGHT ACROSS DIFFERENT PREDICTORS. Some of you suggested that if the b weight was bigger in one equation than another, it meant the relationship was stronger. This is not necessarily the case when the two b weights come from different predictors. Since the predictors are different the units may be different and the comparison does not make sense. In simple regression we only have the different R squareds to tell us about the strength of the relationship. In multiple regression we will contrast BETA weights. Film at 11.

QUOTE OF THE DAY: "The computer is an unforgiving god." -Joseph Campbell