Tom Wenseleers

Linking this post here which may be relevant: https://stats.stackexchange.com/questions/617990/inference-for-high-dimensional-models-based-on-running-a-glm-on-the-union-of-s?noredirect=1#comment1148780_617990 and some quick tests on that idea of doing inference by fitting a regular GLM on the union of selected variables...

Do you think it might also be an option to calculate the variance covariance matrix from the Fisher information matrix calculated for the selected variables, and repeat this procedure for...

@Mamba413 No I haven't seen this exact method being used - perhaps mainly because until recently we didn't have access to good L0 penalized GLM fitting engines.... But it would...

No worries - I have some bioinformatics master students working on this exact topic to figure out what approach would or would not work... I'll let you know once we...

Many thanks for that - that's very helpful! And glad `abess `correctly takes into account the weight matrix for identity link Poisson. For the problem with Gaussian noise I also...

Well I was asking these questions also partly to compare the solution quality of `abess` with that of my own `L0glm `package that I was developing - that one uses...

In this particular example, my variables are not independent, and yet clipping still seemed to perform very well to recover the true solution. In fact, `osqp`, where hard nonnegativity constraints...

Ha that's very cool - many thanks for that! Looking forward to that! Did that result in good performance for this problem then, not only in terms of speed, but...

Many thanks - that sounds really promising - thanks for your work on this! My iterative adaptive ridge method took half a minute for that problem size (currently finishing up...

Many thanks for this! Today I was just testing with the latest version on CRAN (abess_0.4.8) but I am still running into a couple of problems - here a multitask...