statmodel is a collection of Go packages for fitting statistical models.
All results should agree to high precision with standard statistical packages (R, Stata, SAS, etc.). Extensive unit tests against these packages are included in the test suite.
All models can be fit with maximum (or quasi-maximum) likelihood estimation, with optional L1 (Lasso) or L2 (ridge) penalization.
The data provided to the fitting routines are stored column-wise (variable-wise) as either []float64 or []float32 slices. The data type is controlled by the Dtype parameter in the statmodel/core.go file. Note that changing the data type involves recompiling the packages. All calculations are carried out in float64 precision, the option to provide the data as float32 values improves cache performance, and enables analysis with very large data sets.
See the following packages for specific models: