I am currently analyzing tree ring width data with the dplR package. The problem is that my series are very short and I'm not sure how to choose the optimal window.length. My series range from 6 to 27 years. Is there a rule of thumb about how wide the window should be? Is that even possible with this age range?
Do you want to correlate your tree-ring data with other tree rings or climate data? Since your series are very short, it might be difficult to make the correlation analysis "run" through time. Personally, I would not set the window width smaller than 15 years to account for enough degrees of freedom. The window width usually ranges between 15 years to 50 years in dendrochronology.
The size of sliding window has a key influence on the statistical significance. If the window is too small, it is difficult to obtain reliable significance based on T-tests. However, you can try using the methods like Monte carlo or bootstrap that are persuasive even in small samples. Dplr uses the bootstrap method. So even using a small sliding window, even 10 years, it still can give a reliable statistical test and still be convincing.
- Badges
- Science topic
- Similar topics
- Mathematics