v0.1.1

• Override trend / seasonal bandwidth if series small
• StlPlotter#main, StlDecomposition#main input / output CSV via stdin / stdout
• Always use bandwidth corresponding to next odd integer >= n_p for low pass smoothing
• Do not use robustness weights in low pass filter
• Properly pad cycle subseries for low pass filter

v0.1.0

• Use commons-math3 checkstyle, no errors
• Refactor StlDecomposition API (see README.md)
• lowPassFilter uses times in Loess
• Refactored package to com.github.brandtg.stl

v0.0.3

• Uses moving average once in low pass filter, results look more consistent with R
• Set default low pass filter bandwidth to 0.75

v0.0.2

• Default seasonal / trend bandwidth is 1.0
• Weighted mean smoothing of seasonal component if config.isPeriodic() parallels R post seasonal smoothing