- Change precision in output.
- Arbitrary precision for high degree polynomials.
- Automatically choose process
- Linearization - solve all other.
- Fix input handling? (arrow up, sides, don't spam prompt when pasting (or don't have prompt on multiline?))
- Fix NaN on exponential and power regressions (add the min y + 1 to get all numbers above 1, then fit the curve to that)
- Slow ( O(n²) ) Theil-Sen estimator
- O(n log n) Theil-Sen estimator (heavy-hitter, very hard)
- O(n) median (intermediate difficulty)
- All statistical tools for unique lists (not
Clusters) - Performance logging in regression calculations.
- Plotting of data & regressions using poloto or plotlib
- Option for other plot lib.
- Fix bias in power and exponential regressions. Right now, it's biased towards errors in small values, as the large errors are, in the linear space, the log of what they are in reality.
- RANSAC implementation? Iterations = lg(1-chance of success) * (lg(number of data points) / lg(outliers in relation to total))
- Support covariance, for better estimation.
- Non-parametric regression?
- Non-linear regression?