Please could you provide details of how you get standard errors from CV in caret to implement the one standard error rule?

Pretty sure data in some package that is being used was switched form a data.frame to a tibble and something that used to be passed as a vector is now a tibble.....

Hi David,

This is Nima, and I am almost at the end of my PhD program.

I was searching about something related to Introduction to Statistical Learning, and found your work (pdf file and then I used it to find this repo). Such a great work!

I have had a similar idea about this book. My idea was to create an interactive environment so that reader can play around the code and try "what if" questions they have in mind. I recently uploaded three of my notebooks here: blog (it might seem a little bit messy and I may go through it later to refine it)

I am using python and I know it is usually harder to get some statistics (e.g. I think regression in scikit-learn does not provide p-value or at least I couldn't find an easy way for it there, so I had to use statsmodels library. ). My goal is to provide python-code not just for lab/exercise, but for examples of the book and try to reproduce them. I also provided some links to stackoverflow /youtube videos/ ... when I felt reader might get benefit by getting into some details (e.g. I added a link that explains why we need to z-normalize before PCA).

I was wondering if you have any feedback/suggestion about my work. Also, do you think readers of your book will benefit from my notebooks? Because, to be honest, the format you used for writing the book is very nice and clean. And, it would be really great if I could use your insight on cleaning my notebooks and (if you are interested) adding it to the book. For example, maybe just provide a hyperlink that takes reader to a notebook and they can just play with the code. (or maybe try it on their own and then click on it to see the difference and ....)

Please let me know what you think. Sorry for the long post as I got super excited when I found about this repo :)

Maybe hard-code for now. Find better solution for later.

It's in chapter 25. It should be glmnet.

Also, is this Git actively updated?

Gather external references that are usually posted on course website at end of chapters. Also consider ISL and mathematicalmonk videos.

Instead, reference appliedstats. Wait until update to appliedstats.

Create plots similar to: http://www.milanor.net/blog/cross-validation-for-predictive-analytics-using-r/

Change all tables to knitr::kable().

As you continue to add on code that is computationally intensive, you will need to start caching the code chunk results. This in turn means you should start labeling your code chunks (otherwise, they're treated like cattle). Otherwise, we'll have to move you to pushing a rendered product via a netlify setup.

Here you take RMSESD from the results attribute of a train (model) object and then refer to this as se and standard error in the text. However, SD is commonly refers to standard deviation. I think something has to be done to calculate standard error from standard deviation. Or is the output of caret misleading?

Move from caret toward tidymodels org w.r.t parsnips and recipes

Consider looking at Max's Feature Engineering and Selection: A Practical Approach for Predictive Models as well.. Instead of just ISLR.

First, I really like the Bias Variance Tradeoff Simulation Section!! Thank you very much for making such a great book.

One suggestion from me.

Figure 1 is not very intuitive for me, it shows too many information.

Alternatively I think this visualization may be better, which only shows the prediction for the x=0.95 cross different model. From 4 sub plots we can clean see, bias and variance.

Here is the code.

  par(mfrow=c(1,4))
  plot(rep(0.95,n_sims),predictions[,1],col="red", xlim = c(0.75, 1), ylim = c(0, 1.5))
  points(x0, f(x0), col = "black", pch = "x", cex = 2)
  plot(rep(0.95,n_sims),predictions[,2],col="blue", xlim = c(0.75, 1), ylim = c(0, 1.5))
  points(x0, f(x0), col = "black", pch = "x", cex = 2)
  plot(rep(0.95,n_sims),predictions[,3],col="green", xlim = c(0.75, 1), ylim = c(0, 1.5))
  points(x0, f(x0), col = "black", pch = "x", cex = 2)
  plot(rep(0.95,n_sims),predictions[,4],col="orange", xlim = c(0.75, 1), ylim = c(0, 1.5))
  points(x0, f(x0), col = "black", pch = "x", cex = 2)