n1m3 / linregress Goto Github PK

I'm not sure if this could work, but it could imagine that there could be added compile safety for accessing the parameter etc with rustc 1.51?

I'm interesting in running regression on the entire set of variables. However it seems that the FormulaRegressionBuilder requires a formula to work. Is there any thought to adding an "all variables" mode, or else provide an array of variable names so that it's a bit easier to automate? I would like to avoid building a string formula programatically if I can.

Are there any release notes on the changes made from verzion 0.4.4 to 0.5.0? I am upgrading an old codebase that I inherited from another developer and I am encountering some errors that suggest that some of the underlying behavior of linregress has changed in some way as of version 0.5.0. In particular, there was a line that looks like this after upgrading:

Some((model.parameters()[0], model.parameters()[1..].to_vec(), model.se().to_vec()))

This compiles, however now we get a number of unit test failures as a result of this. Would appreciate any insight into what the 0.5.0 changes included and what I can eo to upgrade the code properly.

Just wanted to pop in and say thank you! I'm writing a tutorial with baseball data and wanted a simple, easy to use example. This crate worked perfectly on the first try, due to your excellent documentation.

Thank you!

Not a huge problem, just a bit surprising that the RegressionModel.residuals is returned in the same struct as the others data, but is a different "shape" of data, so when accessed via .pairs() truncates most of the data

For example, the debug output shows this:

    ...
    pvalues: RegressionParameters {
        intercept_value: 5.749549530737332e-269,
        regressor_names: [
            "x1",
            "x2",
        ],
        regressor_values: [
            9.27933982277656e-11,
            0.0008234351722466723,
        ],
    },
    residuals: RegressionParameters {
        intercept_value: -27.015828794746632,
        regressor_names: [
            "x1",
            "x2",
        ],
        regressor_values: [
            -29.59495810026061,
            -26.127909406281596,
            -26.19444350122059,
            -26.362871508164616,
            -26.362871508164616,
            -29.145193769650938,
            -25.75539341699686,
            ...

Accessing pvalues using .pairs() would return the expected data (e.g [("x1", 9.27933982277656e-11"), ("x2", 0.0008234351722466723)])

Problem is doing same for residuals would only return [("x1", -29.59495810026061), ("x2", -26.127909406281596)]

I would like to bring to your attention the fact that your licensing is incompatible with your dependencies.

Your crate is MIT licensed but depends upon nalgebra which is Apache-2.0 licensed only. While Apache-2.0 projects can use MIT licensed components, the reverse is not so.

Consider for instance that MIT is GPL v2 compatible, while Apache-2.0 is not; Your crate being MIT licensed is thus misleading to any potential GPL-v2 projects that consider using your crate - they'd end up constructing a non-compliant product.