I'm verifying the frequency of a certain event over the years. For example, 50% of 50 patients had the event between 1995 and 2000, and I want to compare this frequency in four periods of five years. I have a different number of observations for each period, which is why I'm using frequency,
Should I use chi-square, or a simple Anova test?
Respective to regression, should I use pearson, spearman ou even a logistic regression?
The tests that you mention assume independence of observations. Since you are apparently looking at the same cohort, your observations are not independent. Calculate the autocorrelation function (it may show some seasonality in your data, btw) or try Ljung-Box test of independence. If Ljung-Box is significant, you should look at time-series analyses. AR or ARMA are perhaps the simplest ones, and should work in your situation.
You may start using chi-square for trend: Cochran-Armitage test for linear trend and Pearson chi-square test for any variation.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA