Sen's Trend Test with seasonallity present - Sen T
Another seasonal trend test that has good power detetecting a monotonic trends when seasonal cycles are present. But should only be used when no data are missing. When no data are missing this test is more accurate than the Kendall Seasonal (unless serial dependence is accounted for? See sktt.m for Seasonal Kendall Test and slope.
- Gilbert section 17.4, page 230
There is a subfunction 'rank' in this function that is used to compute ranks for all values in the dataset. Matlab's tiedrank estimates rankings differently then required for this statistic.
Syntax:
[T sig] = SenT(datain, alpha)
inputs:
datain(:,1) = year (e.g. 1999)
datain(:,2) = season (e.g. 1 through 12)
datain(:,3) = values to be analyzed
alpha = for two tail test (e.g. 0.05)
outputs:
T = Sen T value
sig = significance using normal distribution
Requirements:
If alpha is set to zero or not provided, significance will not be
computed. To test for significance, the Statistics Toolbox is
required. Otherwise there are no other functions outside of Matlab itself.
|
