Dear Scholars, I'm working on a trend analysis of hydroclimate time series data. I attempted to compare the magnitude of the trend using Sen's slope and linear regression after smoothing the data noise utilizing exponential smoothing; however, Sen's slope revealed a statistically insignificant trend, whereas linear regression for smooth data produced statistically significant results. Does anyone else have solutions/suggestions for this data smoothing and trend analysis problem?
As a general comment, I do not think you should smooth the data before doing statistical analysis on it.
Anders Buen Thank you for your suggestion. People recommend avoiding data noise when evaluating trends, and Sen's slope is not commonly recommended when there are no outliers. However, my doubt is how the significance level completely changed.
If there is no trend in the iunfiltered data, there are not likely to be a significant trend. How do you know that the filtered data are noise and not real readings?
- Badges
- Science topic