0

Recently I need to test the trend of a time series from 1665 to 1991 and I am going to use the MannKendall test in R. However, I finally found that I only have 210 values instead of the whole 256. There are 46 NAs in my data.

And I just used the MannKendall() in Kendall package in R and got that the P of my data is not significant. I don't know whether with so many NAs I could still just use the MannKendall() to test my data's trend.

Is it still effective? or do I need to transfer my NAs to other values?

Vinícius Félix
  • 103
Michael
  • 13
  • 3

1 Answers1

0

Mann-Kendall test is a test, it tests something, and it does nothing about the missing data. If you are missing data, test cannot do anything about it. What you should see, should be an error message informing you about the missing values present in the data. If you don't see it, the software probably doesn't follow software engineering best practices and instead of failing fast and loud, it silently does something with the data. Here, it just ignored the rows with missing data. If you don't have the data, the test cannot tell you anything about the data.

Tim
  • 138,066