statsmodels

statsmodels → .NET

Verdict: decide. The foundation (linear regression, distributions, basic tests) exists; rich econometrics (detailed GLMs, ARIMA/SARIMAX, mixed models, R-style summaries with p-values and confidence intervals) has no good .NET equivalent. It is a candidate for native code if your usage justifies it.

statsmodels need	.NET
Linear regression, least squares	Math.NET Numerics (Fit, MultipleRegression)
Distributions, basic hypothesis tests	Math.NET (Distributions), Accord.NET
Advanced GLMs, time series, econometric summaries	⚠️ gap — write or work around

using MathNet.Numerics;

// OLS y = a + b·x
(double a, double b) = Fit.Line(xs, ys);
double r2 = GoodnessOfFit.RSquared(xs.Select(x => a + b * x), ys);

Pitfalls

• No rich summary(). Standard errors, confidence intervals, coefficient p-values are not provided out of the box: compute them yourself (the estimator covariance matrix) or port them.
• Time series. Nothing equivalent to SARIMAX/statespace: either restrict the scope, or make it a native lot of its own.

Before any native development here, weigh the real need: a regression plus a few tests is often enough, and Math.NET already covers that.

Guide to be expanded as real needs arise.