choosing the right non-parametric test

#1
Hi there, I am trying to analyze some data and am having trouble trying to figure out what to do. The data is a shrub/elk browse study. The data I want to analyze is:

Year: 2001 and 2006 (two years the study was done)

Transect: (37 transects total, replicated both years)

Plot: There are 50 plots or measuring points on each transect. At each plot, the nearest shrub to the observer was recorded. We were looking at three different types of shrubs, so there are 3 different species categories

Species: 3 species of shrubs

Form class: This is a measurement of how much the deer and elk have eaten the plant. It is made up of 3 classes (1-lightly browsed, 2-moderately browsed, 3-severely browsed)

This is paired data in the fact that it was done at the same transects and same number of plots both years, but it is different in the fact that you may not record the same shrub at each plot year to year.

I want to show analyze the difference in the shrub classes by species between the two years. So I will want to group the data by species so the analysis is done for each of the three species.

I did the simple stats and there has been a HUGE decrease in the amount of lightly browsed species (for all three) and a HUGE increase in the amount of severely browsed shrubs (for all three).

Any suggestions on the best test to use for this data. Originally I thought about the kruskal wallis but it will only test the change in the numbers between the two years, not the change by each form class. Chi-squared maybe??

Thanks in advance!
 

JohnM

TS Contributor
#2
Chi-square sounds good to me, since you're comparing two years and you have a response variable that is categorical/ordinal.