Without getting too deep into the specifics of the project, I am examining gridded precipitation data across two areas - one to the east of a body of water, and one to the west - to see if there are significant differences between one side and the other for a series of precipitation events. One side has about 720 gridpoints, and the other has about 620 gridpoints.
The problem I'm having is that in most of the events, there are a significant number of 0's (indicating that cell did not receive any precipitation), a few hundred in many cases, and I'm concerned that something like a t-test will prove insufficient. Am I right in that concern? If so, what are some options for me to consider as an alternative means of analysis? Any help would be MUCH appreciated!
As I've gotten deeper into this project I've realized just how rusty I am with regards to statistics - it's been a few years since I've really done much :)
https://academic.oup.com/bioinformatics/article/31/14/2310/255807
May read it.
You can replace the zeros with NaN. Then it will not participate in the statistical analysis.
Maybe you can generate two subsets of data with all grids with zeros discarded, and then make statistics on the subsets. Zeros means no precipitation, thus are meaningless for the statistics.
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology