Hello, I desperately need some help with my project...
If someone could help me, I would be very very grateful!
I am comparing the different results obtained through different survey methods (visual and acoustic surveys) to study whale distribution. I have data from three different cruises (A=summer, B=autumn and C=winter) in the same area.
The study area was divided into a grid of about 300 cells, and for each of them it was calculated an Encounter Rate (ER).
For each cruise, I have two different ER: one based on the visual survey, and one based on the acoustic survey.
I would like to test:
- For each cruise, are there significant differences in the ER among visual and acoustic surveys?
- For each survey methodology, are there any differences in the ER among seasons (cruises)?
- ER are not normally distributed, and most of the values (more than 95%) are zero.
- I have different sample sizes (each cruise and each method do not cover ALL the study area, but for each of them I have a different number of cells, which varies between 75 and 261)
What should I do? I wanted to use a non-parametric test, but which one?
Remember that more than 95% of the values are zero...
Is there a way to deal with this stuff in a statistically acceptable way?
There are a couple of folks here with a lot of expertise in working with biological and ecological data who may have some ideas - but my first thought is that you could be looking for something like either Poisson or negative binomial regression. Both of these analysis methods are designed for use with count data - it sounds like your Encounter Rate variable is a count variable? The encounter rate (by cell) could then be your response variable, and the variables "season" and "method" could be categorical predictor variables.