I am using the Kendall function in R to do the correlation between (Vi, Ni) where Vi is a continuous variable (volume of stream flow in a reservoir) and Ni is a discrete/count variable (frequency of the stream flow above a defined threshold).
How should we interpret the below correlation results between (Vi, Ni)?
tau = 0.950, p = 0.546, n = 17
Is it "there is no significant positive correlation between two variables even though the tau value is very high?” I am not sure how to interpret this when the tau value is very high.
p-value that u have observed insignificant may be small sample size.
Some possible reasons are:
1. Your sample size is small. The critical value for this test, like most statistical test, largely depends of sample size. The smaller the sample size, the larger the critical value.
2. Your data set may have many ties.
3. The unique values in the ordinal variable might be very limited. If it is a binomial variable, using bi-serial correlation can solve the problem.
Another option that can fit your goal is using logistic regression.
Regards,
The result doesn't sound right to me.
What is the "Kendall function" you are using?
Would you be willing to post your data or toy data that is giving a similar result?
Thank you all for your responses and suggestions. Unfortunately, I have maximum 17 years data for streamflow in the reservoir. The main goal is to see whether we can find a significant relationship between the streamflow in the reservoir and frequency of stream flow in 110 stations in the study region.
Dear Hormoz Sohrabi, yes there is lots of ties data with zero values. Thanks for your suggestions on the bi-serial correlation and the logistic regression. I will take a look at theses methods.
Thank you everyone. I just wanted to make an update here. I checked all steps of my code and it was correct. When I have gone through the stations, where I have seen high p-value and high tau value, I found that those stations are having many zero data ( about 90-95 percent). Thanks again everyone for taking your time to help me.
Hi Maryam Pournasiri Poshtiri, I just noticed this same issue in my results. How did you work around these stations? You used logit regression or bi-series correlation? Let me know, please. Thanks
Hi Oghenekevwe Christopher Oghenechovwen ,
As I said, those are stations with mainly zero data. I used Kendall correlation and just marked those stations as zero-flow stations.
Hope this helps.
Yes it does! Many thanks Maryam Pournasiri Poshtiri. You rock!
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation