New answers tagged generalized-linear-model
1
vote
Accepted
Calculate log-likelihood of logistic regression
A likelihood is a strange thing: it is not a probability and does not need to sum or integrate to $1$. In fact likelihood is only measured up to proportionality, with a positive multiplicative ...
- 35.1k
7
votes
What is the adequate regression model for bounded, continuous but poisson-like data?
Consider ordinal regression. You have data that are ordered but, from your description, it doesn't seem that the difference between scores of 1 and 2 is the same as the difference between scores of, ...
- 81.2k
0
votes
Is there a way to correct for degrees of freedom when using a generalized linear model with a Poisson distribution featuring random effects?
I had the same question. My understanding is:
when the distribution linked to your test-statistics (e.g. z-value, t-value, chi-square, incidence rate ratios) depends on the number of degrees of ...
- 687
0
votes
How to use Logistic Regression/Decision Tree with 'inflated' responses
This is a common occurrence in machine learning and exactly the issue of unbalanced-classes that gets discussed on here so often. The typical recommendation is to do nothing, as class imbalance is ...
- 50.6k
0
votes
Using ANOVA for variable selection in Multinomial Logistic Regression
This response is based on my experience working with colleagues who have made similar analysis suggestions. The rationale behind their suggestions (and what I am assuming is the same for your advisor)...
- 4,782
5
votes
Accepted
Manually calculate the variance-covariance matrix for a fitted GLS model -- i.e., vcov(glsModel)
In an attempt to match your variable names, suppose the (conditional) error covariance matrix $\Psi = \sigma^2 \Sigma^{-1} = \sigma^2V$ in your general linear regression model is know up to a positive ...
- 4,440
8
votes
What test does summary() perform on a glm() model using a Gamma distribution in r?
Demetri Pananos makes the critical point about regression models (+1): the coefficient estimates are taken to have underlying multivariate normal distributions, at least in the asymptotic limit of ...
- 81.2k
8
votes
Accepted
What test does summary() perform on a glm() model using a Gamma distribution in r?
But a Wald test is a parametric test which assumes a normal distribution. Is R perhaps performing a Wald Log-Linear Chi-Square Test instead of a normal Wald test in these cases?
The coefficients are ...
- 31.3k
0
votes
Algorithm selection rationale (Random Forest vs Logistic Regression vs SVM)
Logistic regression was invented by a statistician, for statisticians. SVMs are a true ML algorithm. Random Forests are a statistician's take on Machine Learning.
Since you explicitly ask about "...
- 7,969
1
vote
Accepted
Can glm(m) model estimates be negative after back transforming?
Assuming your Poisson and negative binomial models used the (natural) log as the link function, yes, you would transform those coefficients by exponentiating them. It is certainly possible to have a '...
0
votes
Generalised Linear Model (GLM) linearity assumption checking
It is best not to use those plots for non-OLS regression models (cf.: Interpretation of plot (glm.model)). If you want a test of the assumption of linearity (on the transformed scale), you can add ...
2
votes
Interpretation of quadratic term in Poisson regression
You can assemble two useful pieces of information to arrive at the interpretation of the linear and quadratic term in a Poisson GLM:
As you already stated, $\exp(\beta)$ for any first-order ...
- 57.8k
0
votes
Accepted
Modeling simple longitudinal data, unknown trend
What you describe is very similar to a MMRM, which would additionally account for the correlation of measurements over time (and do some implicit imputation of missing timepoints, which you would ...
- 28.2k
Top 50 recent answers are included