Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



correlation
Posted:
Oct 3, 1996 2:48 PM


[I tried to post this earlier & think I failed; my apologies if I'm repeating it. The original query sought a good exercise on probability, and various responses came in with suggestions for problems or data to use.]
> To any of the other fine suggestions that responded to this post, add a > segment on probability. Confine the data to 10 cases, and calculate r. > > Next, arrange a deck of cards so that you have only the numberd hearts > (ace10) and numbered spades. Shuffle the cards and deal them out. Ask > what you would expect  best guess  the correlation to be between the > values of the two suits. Now calculate it. Do it a couple of times. > > To move things more quickly do it on the computer. Here are Resampling > Stats commands; I'm sure this could be done in other packages as well: > > DATA 1,10 hearts > DATA 1,10 spades > SHUFFLE hearts hearts* > CORR hearts* spades r > PRINT r > > After doing it a number of times, ask students to explain how the cards can > produce substantial positive and negative correlations (10% of the results > will be >.5 or < .5). > > Now do the cards example lots of times. > > DATA 1,10 hearts > DATA 1,10 spades > REPEAT 1000 > SHUFFLE hearts hearts* > CORR hearts* spades r > SCORE r z > END > HISTOGRAM z > > Observe the distribution of the r produced by random shufflings. > > Ask them to interpret the observed value in light of these data. > > (note: it helps to have the 10 real cases with observed "r" is between .3 > and .45  big enough to suggest a relationship, but small enough to come > within the range of chance variation) > > > Peter Bruce > > > At 04:46 PM 9/30/96 0400, you wrote: > >Note: this is a serious request, and not a joke. > > > >If you were informed that a ranking official from the federal government > >was going to visit your AP Stat class while you were doing a unit on > >correlation, what activity would you do? All suggestions welcome, and any > >one I use will receive proper credit refernces. Thanks in advance. > > > >Chuck Peter Bruce voice 7035222713 Resampling Stats fax 7035225846 612 N. Jackson St. stats@resample.com Arlington, VA 22201 www.statistics.com



