# Thread: Is t-test appropriate for comparing percentages?

1. ## Is t-test appropriate for comparing percentages?

I am trying to analyze a seed germination trial comparing 2 different germination treatments. I sowed seeds from 10 individual plants in 10 different petri dishes for each of the 2 treatments, with between 22 and 200 seeds per dish, and recorded germination rates (number of seeds germinated per total number sown). Individual plants are not paired across treatments. I did a t-test comparing germination rates for the 2 treatments, with n=10 for each treatment. I am wondering if I should do a different test, as germination may also be looked at as binary data, with seeds either “germinated” or “not germinated.” If I had run the trial for a single plant, I would have done a chi-square test. But because I have n = 10 plants, I’m not sure how to do this. I greatly appreciate any input!!!

2. ## Re: Is t-test appropriate for comparing percentages?

You have to test the percentage germinated on hypergeometric distribution (on binomial distribution, if a germinated seed can germinate a second time by cutting for instance), I think.

 Tweet