Academe‎ > ‎eTranscript (UAA)‎ > ‎

Statistics for Psychology (Fall 2012)

Reflection

Below is my work space in preparation for the first examination which consisted of three parts: short answer, multiple choice and computation.


Here is an email I am very proud of sent by the professor in response to a question I asked:

On Nov 19, 2012, at 12:02 PM, John M Petraitis wrote:

Ezra,

I'm not sure I've ever had another student in the past 20 yrs dive so far into the details. Not only did you wonder about a quirkly effect of huge samples sizes (that they increase power, but also lead us to say there is some effect when the effect is really small...but bigger than RSE), you looked up power curves. Nicely done.

Connecting your original question to today's lecture...If you have a huge sample size, you might make a black/white decision to reject Ho (arguing that there is some effect that is bigger than RSE), but you might still have a very minor effect size (a Cohen's d around .10). Big sample sizes decrease RSE and increase power, but they do no increase the effect size. So, really big samples can find really, really small (perhaps unimportant) effects.


John Petraitis, PhD
Professor of Psychology
Associate Dean for Social Sciences
College of Arts and Sciences
University of Alaska Anchorage