statistics
statistics copied to clipboard
haskell
A fast, high quality library for computing with statistics in Haskell.
A [recent built on the nixos build server hydra](https://hydra.nixos.org/build/122996414/nixlog/1) failed with this error for `statistics-0.15.2.0`: ``` Tests for: BinomialDistribution log probabilty check: FAIL *** Failed! Falsified (after 34 tests): binomial...
I have been porting part of this library to Idris and so I have been looking at it with more precise types in mind. Which leads me to: [`solve ::...
I sometimes see the Spearman correlation test fail in the test suite. Example: ``` Spearman correlation, nonlinear: FAIL *** Failed! Falsified (after 3 tests and 7 shrinks): [(-0.58,0.0),(-0.4,0.0),(-0.304,-1.0)] [(-0.58,0.0),(-0.4,0.0),(-0.304,-1.0)] [(724.6376811594204,180.0),(711.5942028985509,180.0),(704.6376811594205,200.0)]...
I occasionally see the testsuite fail for the gamma distribution. ``` log density check: FAIL *** Failed! Falsified (after 54 tests and 2 shrinks): improperGammaDistr 80.10044478990355 0.5569517767725765 36.1 density =...
Practically all statistic packages provide a p-value together with the Pearson or Spearman correlation coefficients. Can we also have this feature? I found https://github.com/haskell/statistics/pull/125, but it was never merged. EDIT:...
A colleage tells me this is commonly done. I think scipy does the same: https://docs.scipy.org/doc/scipy/reference/generated/scipy.stats.skew.html Sorry I don't know very much about this subject, but wanted to make an issue...
As a non-statistician, I need to run linear regression on a set of data points, such as: ``` measurements = [(1.0, 2.0), (2.0, 2.5), (3.0, 3.0)} :: [(Double, Double)] --...
This is one way to resolve #146. All examples here will use `mean` but readily generalizes to all fold based estimators and with some problems to quantiles and like which...
Test for quantile inverse fails spectacularly: ``` Quantile is CDF inverse: FAIL *** Failed! Falsified (after 99 tests): improperBetaDistr 1.9354985700057857e-2 9.298452506189731 Double01 0.7873411995889938 Quantile = 9.435096862311606e-5 Probability = 0.7873411995889938 Probability'...
So far following problems with Student-T distribution were identified: 1. `quantile` lose precision for pā0.5 since x becomes close to 1, while p is near zero ```haskell quantile :: StudentT...