I have been recently reading about the advantages of several multivariate regressions with penalization for high-dimensional data (for example, gene expression analysis associated with prognosis). When do you think that penalized models might also be useful in practical terms for clinical data in which the number of potential predictors is usually shorter?
I think they are useful even for any data. I have seen recommendations from some authors about their usage.
The motivation for penalization procedures in multivariate regression (e.g. LASSO) is to avoid overfitting (i.e. if the number of predictors exceeds the number of samples). However, if the number of predictors is much smaller than the number of samples, penalization procedures might still be helpful since they can potentially eliminate meaningless predictors.
And what do you think of a database with >300 rows and less than 10 predictors after univariate screening analyses. This is a typical situation. Still useful?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression