Hello all,

I'm trying to figure out which non-parametric test to use for my data, and I'd really appreciate your input. In my study, I had subjects rank 12 sets of 3 objects on some property. (So, there were 12 trials)

For example (and this is not the actual experiment, but it's a handy analogy): for each set, I present them with 3 foods and ask them "out of these 3 foods, rank which one is healthiest, 2nd healthiest, and 3rd healthiest". Each of the 12 sets contained 1 fruit, 1 vegetable, and 1 junk-food product.

I would to test whether there are any significant differences in the "heath" rankings between the categories of food.

Numerically speaking, I found that the "Fruit" and "Vegetable" categories were ranked as "most healthy" about equally over the 12 trials, with Vegetables having a slight advantage: Vegetables were ranked as most healthy 55% of the time, with Fruits being ranked most healthy 43% of the time. Junk-food was only ranked as healthiest 2% of the time.

While it's clear that Junk-Food was not considered healthy, I would like to quantify this statistically. As well, I am interested in determining whether the numerical difference between the Vegetable and Fruit rankings is statistically significant.

Can anyone suggest a test to use? Thanks in advance!