A personal blog about my fumblings with statistics, finance and anything R


Built using jekyll and hyde

Clustering

Here’s an example of how one could use 3d plots and some clustering to make some inferences from available data. I am using a set of financial indicators available on FRED. You can download it here. I wanted to compare two different time periods using this data:

Sparse Group Lasso

As a follow up to my last post, here’s a post on Sparse Group Lasso.

Group Lasso

There is a nice extention to the Lasso which lets variable selection work on a group of variables. Hence, instead of a single variable entering the mix, an entire group of variables enter the regression equation together (see Yuan and Lin). Also see link

Lasso and Ridge Regression

I wanted to follow up on my last post with a post on using Ridge and Lasso regression . Lasso can also be used for variable selection.

Bootstrapping StepAIC()

I work in the field of finance and find that people often rely on OLS regressions for doing predictive analysis. It’s easy to implement and everyone knows about it. OLS regression gives us a very well developed mathematical framework which can be used to develop linear relationships. These relationships can then be used to create forward looking projections. But what about the features? How does one go about selecting the right feature set which can be used to reliably predict the variable under consideration?