1
$\begingroup$

I want to test the correlation between variables considering several study areas. I aim to study habitat selection using telemetry data (locations) in five study areas, and I have variables regarding landcover (10; in % for each study area), topography (1) and distance to water/human settlements (2; mean for each study area). In total, I have 13 variables Before starting with habitat selection analyses, I want to check if the variables I am using are correlated, in order to perform analyses using only independent variables. I am not sure how to analyse this with several study areas.

What is the best way to perform the analysis?

Improve this question
$\endgroup$
9
  • 1
    $\begingroup$ Can you say more about your situation, the variables, & your goals here? This is rather sparse & abstract at the moment. $\endgroup$
    – gung - Reinstate Monica
    Commented May 29, 2016 at 17:42
  • $\begingroup$ I just updated the post, if any more information is necessary please let me know! $\endgroup$
    – mto23
    Commented May 29, 2016 at 17:59
  • $\begingroup$ So you will have a dataset with 5 rows (for the values from each location), & 1 column for each variable, is that right? How many variables are there, 3? $\endgroup$
    – gung - Reinstate Monica
    Commented May 29, 2016 at 18:02
  • $\begingroup$ Yes, 5 rows (one for each study area), and 10 variables regarding land cover, one related with topography and the two distances (to water and to human settlements); so 13 variables in total. Within each study area I have several animals with several locations, but I don't think that should be important for the correlation between the variables, or should it? $\endgroup$
    – mto23
    Commented May 29, 2016 at 18:19
  • 2
    $\begingroup$ Update your question with the numbers of variables. You have a matrix of [5×13] and you want to find the correlation between these 13 variables based on the 5 observations? There are too few data to do such a work. If you obtain more data, this question has been asked and answered several times in Stackexchange. There are techniques like: principal component analysis, pearson correlation, ... $\endgroup$
    – PM0087
    Commented May 29, 2016 at 18:34

1 Answer 1

Reset to default
0
$\begingroup$

There are several techniques known in "parameter selection" section part of a work. After parameter selection comes modelling.

In parameter selection you analyze relations between parameters and rank the parameters in order of importance. Here are some of the works:

  1. A Regression model and observe the p-values of the coefficients of each variable
  2. Pearson Correlation
  3. Spearman Correlation
  4. Kendall Correlation
  5. Mutual Information
  6. RReliefF algorithm
  7. Decision trees
  8. Principal Component Analysis, etc.

Note that you need enough amount of data for your results to be accurate.

After you have performed these variable importance analyses, you can rank the variables and then compare the ranks and thus the case studies.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.