I have a large data set of admission and presentations over time. I am wanting to compare the admission rates of several time periods.
How is best to go about doing this?
I have been using the Poisson's means test to two groups at time but is there a statistics that can do it all at once?
I think you can use the 2-Sample Poisson Rate Test with Minitab Software.
If you are using time as a predictor variable, then it makes little sense to test pairs of time points. You can model changes over time using a linear model. A straight line really only answers the question of whether they have changed over the time period. To get the shape of the relationship, you might look at fractional polynomial regression, which is capable of capturing odd-shaped linear relationships. And remember to use robust variance estimation, and check your model fit.
I also think that if you don't have naturally given time intervals you should better use time as a continuous predictor variable, presumably in a binomial model.
- Badges
- Science topic