Biomedical Education Poster Session

Joseph Baker
jbaker@hms.uq.edu.au

While regression analysis is indeed a measure of association, its use as a predictive tool in the social sciences is not without precedent.  In fact, independent variables in social science research are commonly known as "predictor variables" in regression analysis.

Rather than continue our discussion of the semantics of the word "predictive" and my "far fetched" results, I would draw your attention to a number of texts written by persons much more learned on the subject than I.  

Harrison, and Tamaschke (1984) Applied Statistical Analysis.  Prentice-Hall. Pgs 317-320.  

Williams (1979).  Reasoning with Statistics.  Holt-Rinehart-Winston. Pgs. 137-147.

My apologies for not having more up to date references.  These just happened to be two texts that were within reach.

On Mon Dec 7, grover wrote
--------------------------
>Dr. Baker:  I disagree with you in that correlation and regression analysis are predictive.  Mind you I am not questioning their usefulness as a tool to understand potential mechanisms but merely questioning the use of an associative tool for its predicability.  Essentially, for a predictive measure, if you were given the known variable, would you be able to predict the outcome.

>On Sun Dec 6, Joseph Baker wrote
>--------------------------------
>>Dear Dr. Grover,

>>Thanks for the comments on the study.  I hope you are enjoying the conference and I would enjoy hearing your suggestions for reasons why the results came out as they did.  I must disagree with your suggestion that correlation and regression analyses are not predictive.  Don't both these analysis techniques provide a measure of how well an outcome (i.e., binge drinking) can be predicted based on knowledge of an independent variable (sensation seeking or general self-efficacy)?  As to your comment about peer pressure, I whole heartedly agree.  That is likely a variable of significant influence.  However, regardless of peer pressure, the moderating relationship for self-efficacy held for marijuana use and sensation-seeking only suggesting that there is something unique about this relationship.  Certainly this area needs further examination.

>>On Fri Dec 4, grover wrote
>>--------------------------
>>>Dr. Baker: I am glad to see your nice presentation and hope you enjoy the meeting.  I found your conclusions on predictability a little far fetched.  The last time checked, correlation was not a predictor.  It is hard to say whether the binge drinking affects self efficacy or vice-versa.  Also, there is big parameter missing in this analysis - peer pressure.
>>>
>>
>