>I haven't used m & ms at the beginning of the year, but last year I >used them when we began hypothesis-testing, first to test whether the >proportion of blue ones was the same as the proportion of the tan ones > they replaced (using the proportion of tans given by Mars and >published in Moore and McCabe), then to introduce Chi-Square >goodness-of-fit tests (for the distribution of all colors). We found >that there was *enormous* variation in proportions of each color >between manufacturing lots. So if you buy all your bags from the >same carton in the same store, you might get tons of yellows and >almost no greens. The same goes for buying one 2-pound bag. So you >have to get your students to spread out over town and buy bags of m & >ms from a "random" selection of stores! >
sounds like everyone is just working with the wrong model.