I have a digital elevation model of a study site. The study site (5 ha) has a contrasting relief, that is, it has very low and very high elevations. I used a GIS software to extract the elevation data out of 500 randomly distributed points.
Question 1: Should I expect the elevation data to follow a normally distributed curve (normal distribution assumption) so I can use parametric statistics?
Question 2: What specific technique (author or reference) should I use to remove outliers so the elevation data fit the normality distribution requirement (parametric statistics)?
Thanks.
I assume you get elavation data like in this video:
https://www.youtube.com/watch?v=isDop9WKdjk
The data are in the column of an Excel file. They are continuous.
To see if these follow the normal distribution, you can run the Shapiro tets and inspect the histogram. If this distribution is fairly normal, just run a parametric test, if not you can use an non-parametric test such as the Kruskal-Wallis test.
Note that the t-test and analysis of variance are quite robust tests. If you have a large number of data, like you seem to have since there are 500 randomly ditributed point, just do not wurry about normality of distribution and run a parametric test.
I wouldn't think that elevations would tend to follow any particular distribution. I can imagine a landscape with flat valleys and flat plateaus that might have a very bimodal distribution. On the other hand, rolling hills of a certain shape might have a normal distribution. There's a histogram presented here: https://rpubs.com/GeranMehr/639487 .
There's no harm in plotting a histogram of the data and seeing what it looks like.
You shouldn't exclude outliers unless you know them to be incorrect values. I would think that published DEM values would have gone through some quality control, and so should have any outrageous values.
You might look into generalized linear models, maybe Gamma regression, for how you might want to handle your analyses.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python