lambda in smoother matrix (chap5, 5)

[lzhang12]

Firstly, thank you for your nice work. I am working on my own code to reproduce the examples in the book, and came across a problem for the spinal BMD data shown as Figure 5.6 in the book, and worked out in your notebook "smoothing splines.ipynb" of chapter 5.

Basically, I can reproduce the results, the only issue is the choice of lambda, the regularization/smoothing parameter. In the book, it uses 0.00022, but in my code, this value has a much smaller smoothing, and leads to a large overfitting.

And I see in your note book, you multiply a 3600 factor before lambda. With this, I can fully reproduce the smoothing curve in the book. My question is where is this 3600 coming from ?