Documentation for splines could be improved

Hi

this is more a request/proposal than a question. It turned out yesterday when I asked about natural spline, that natural splines, as John pointed out, are supported by matlab, howver the documentation is not very helpful.

Natural splines are the first to be mentioned in many textbook, so help spline should say a word about it, it should also provide a link to csape, but it provides a link to splines, which provides then information about csape, but this is not enough. Maybe somebody of the Mathwork teams reads this and take notice.

thanks

Uwe Brauer

[John D'Errico]

As I argued at length in my response on your last question, I'll argue that making it easier to gain a natural cubic spline fit is not necessarily a good thing at all.

https://www.mathworks.com/matlabcentral/answers/325706-natural-spline-unmkpp-gives-wrong-answer

Were that to happen, then we would find too many cases of people having problems with the spline fits they get, and often not even knowing they were using the wrong tool to solve their problems.

So, it is indeed possible to gain a natural spline fit to your data, and you can find it, IF you look. I showed how to do such a fit in two distinct ways using the existing spline tools.

But making it more obvious and easier yet to build a natural cubic spline will only give novices (who do not understand the issues and the reasons to avoid them) the wrong tool to use, and the comfort of thinking they are just using what they were taught in some long forgotten class.

I'll argue that this is the same issue as with det. Det is perhaps the most misused tool in MATLAB (with inv in second place.) That happens because it is there, easy to find. Students are taught to use a determinant to test for singularity of a matrix. But those same students are rarely taught why they should avoid using det after all, that other tools are a better choice. My point is, had det been provided only through some less obvious source, we would have fewer problems. Yes, det is here to stay, as is inv, but lets not create another det.

[Uwe Brauer]

I agree with you on det and on inv, as a matter of fact, I 'forbid' my students to use these commands at all costs.

However with natural splines it is a bit different. In a course on advanced numerical analysis in which interpolation and friends are discussed you are completely right, but not for a beginner course.

Because of time constraints it seems to me not feasible to discuss the benefits of more sophisticated and more difficult to gasp splines. One could easily add a sentence in the documentation about natural splines and that there are crude tools and should be used with care or best be replaced with more sophisticated tools.

But don't make it difficult for the user to find them, it took me almost an afternoon to find it and I was already thinking of writing my own small routine, which right now I have to really the time to do.

[John D'Errico]

I do admit this is only my opinion. And, yes, one could add a line or two indicating where one might find a natural spline. It would not hurt, especially if a note of caution was added in the same spot. I'm a bit gun shy, mainly due to det and inv (both of these tools do have some valid uses.) I honestly don't think that Cleve (and company) envisioned the many problems novices would have over the years, just because they find det so easily. Even more people never notice a problem, blithely using the wrong tool.

[Stephen23]

"especially if a note of caution was added in the same spot"

The warnings (help and mlint) do not stop beginners using inv though.

Answer 1

[Image Analyst]

At the bottom of the help, when it asks "Was this topic helpful?" click NO and fill out the form. I know for a fact that actual real live people read what you submit.