# Comparing proportions

#### bvelten

##### New Member
Hello,

I need some help knowing what stats to run for my data analysis.

I have samples of muscle collected during different seasons and I am analyzing the different types of a specific protein present in the muscle. I want to compare across the different seasons to see if there is a statistical significance in the relative proportion of the protein types.

I have a sample size of 6-9 individuals and samples from each individual are run in triplicate. So I have 18-27 values which are used to produce an average for each season of the relative proportion of each type.

So for example, summer muscle samples have an average of 35 +/- 9% Protein Type A and 65 +/- 9% Protein Type B. Autumn muscle samples have 55 +/- 16% Protein Type A and 45 +/- 16% Protein Type B. So I want to compare the average proportion of the protein types across the two seasons to test if there is a statistically significant change in protein expression.

I have read up on using Fisher's exact test because my sample size is smallish. But I'm not sure it's appropriate in my scenario. Any help is greatly appreciated.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Fisher's would not be used with means here.

If you have two groups (i.e., season) you would want to examine if t-test or Wilcoxon rank sum may be appropriate.

#### bvelten

##### New Member
Thank you! I had originally proposed using a t-test, but my advisor suggested that it wasn't appropriate because the values are dependent on each other (i.e. the proportion that isn't Type A is made up of Type B), so he suggested a Chi-squared test.

I actually have three groups (winter, spring, summer) total and want to compare whether there is differences across all three, but I'm assuming I can test winter vs. spring, winter vs. summer, spring vs. summer.

#### hlsmith

##### Less is more. Stay pure. Stay poor.
You are then looking at either ANOVA or Kruskal Wallis with follow-up pairwise comparison if they are significant. You can compare all of the groups, you just need to correct your level of significance (alpha), so you don't accidently find something to be significant by chance.

#### bvelten

##### New Member
Can you explain further what you mean by correcting the level of significance?

#### hlsmith

##### Less is more. Stay pure. Stay poor.
Concept: false discovery rate or multiplicity.