I am looking for statistical methods used to compare frequency of observations between two groups. suppose I have two groups of pigeons with the same sample sizes, say 10. The number of red blood cells of any pigeon are counted. The observations are as follows:
Group 1: 66, 58, 64,48,58,65,52,57,54,58
Group 2: 66, 59, 64,58,62,59,57,66,54,58.
Can you suggest a statistical test to compare frequency of red blood cells in two groups?
It seems you are accidentally complicating your approach. Usually, "frequency" analyses refer to counting how many times out of the 10 sample you found (yes) or did not find (no) a particular observation. Instead, you are actually counting the number of red blood cells and trying to compare this number between two groups.
Thus, you can simply use a t-test as long as (1) the data from Group 1 are completely independent from the data in Group 2 (they would not be independent if, for example, you measured the same pigeon twice and Group 1 was 'before' a treatment and Group 2 was 'after' a treatment. If that were the case, you would need to use a paired t-test), (2) the data is normally distributed, and (3) data is randomly sampled from the population.
Sometimes, the type of data you are measuring might be referred to as "count data", but that terminology is mostly used to describe data that tends to be zero-inflated. This is because when you are counting, zero is typically the minimum value. Your data is of course not zero-inflated at all, but you should still test for normality before using a t-test to compare the number of red blood cells in each group.
Good luck with your analysis.
t-test would be appropriate to compare the means and standard deviations of counts for your two groups. Two-group ANOVA is mathemetically equivalent to the t-test.
A really general test you might use is Kolmogorov-Smirnov two sample test. It does not assume anything about distribution or variances, ect... It is non-parametric. It compare if the two distributions are coming from a common distribution.
Such a test is pretty much universal.
I would would fit Binomial or Kernal distributions on the data and compare the fitted distributions.
good luck!!
Dear Kambiz
You must test your data for normal distribution and independence between groups, in case the assumptions are available, then you can use T-test for two independent samples.
But in case of not normal distribution, you can use nonparametric test ( Mann-Whitney)
Good Luck
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis