Basic, technical questions on “goodness-of-fit” for regression models… Using N data points, suppose I fit a regression model with n_\beta parameters \beta to get a predictor \hat{y}_i = \hat{\beta}\phi_i, with prediction error e_i = y_i - \hat{y}_i. The standard deviation \sigma_e is then given by…

If this is a Julia question, it’s not very clear what the question is :slight_smile: if it’s a stats question, maybe that https://stats.stackexchange.com is a better location to ask such questions. If you have specific questions about a given package like GLM then please link to the part of the co…

I apologize for being vague in my question… Essentially, what I’m after is the following. Suppose I have a number of candidate regression models, and I want to find which one is best. The models could be generated by ANNs (e.g., a single hidden layer with varying number of nodes), PCR, PLS, GLM, etc…

As a statistician who uses Julia I will comment, perhaps at too great a length, on the second question. A linear regression model is closely linked to the Euclidean geometry of the N-dimensional response space, because the assumptions on the distribution of the response - independent, constant vari…

For the model shootout, you could use MLJ however it’s probably not going to be super simple because here you will potentially want your own custom models. But, to simplify, let’s say you only consider models that are already provided like, a bunch of linear regression models, a decision tree or wha…

[image] BLI: what would be a useful measure of “goodness-of-fit” This really depends on your application (ie what aspect of the data you care about). See eg the references in [Amelio Vazquez-Reina] What are posterior predictive checks and what makes them useful? ba…

I’m not well versed in statistics… but " the posterior predictive distribution is the distribution of possible unobserved values conditional on the observed values"… sounds like computing the “goodness-of-fit” from the training data while also taking into account the distribution in estimated parame…

It can be thought of as related, but the important property of Bayesian posterior predictive checks from the point of view of this topic is that they impose a metric (you evaluate whatever is important for your purposes) and at the same time put a posterior distribution on it, so you can think about…

Regression and goodness-of-fit?

New to Julia

tlienart February 27, 2021, 9:56pm 6

By the way there’s a similar topic being discussed in a separate thread with pretty good answers by @rikh and others Bayesian Model Selection - #2 by rikh

Topic		Replies	Views
LinearRegressionKit results correctness Statistics	0	488	June 30, 2022
GLM is slow on large datasets. Using OnlineStats for regressions? MixedModels? Performance glm	25	5413	November 26, 2018
Fitting generates an extreme standard deviation (Dataframes + Easyfit) General Usage	2	134	September 27, 2025
[ANN] RobustModels: Robust linear regression using M-Estimators and more Package Announcements statistics	7	1897	March 20, 2023
MLJ Tuning and Hyperparameters , Regression Performance optimization , machine-learning , mlj	0	305	November 20, 2022

Regression and goodness-of-fit?

Related topics