Clarification needed

This task needs more clarification, like a link to a suitable wikipedia page. —Donal Fellows 17:04, 29 June 2009 (UTC)

This task requires merging with Polynomial Fitting already representing least squares approximation example in the basis {1, x, x²}. Linear regression is just same in the basis {1, x}. --Dmitry-kazakov 18:31, 29 June 2009 (UTC)

OK, there's now implementations in two languages. It's not clear to me how this is different from the polynomial fitting task either, but I'm a completionist (for Tcl) and not a statistician... —Donal Fellows 10:34, 9 July 2009 (UTC)

An explanation from the Lapack documentation may be helpful. [1] The idea is that you want to model a set of empirical data

$\{(x_{1},F(x_{1}))\dots (x_{m},F(x_{m}))\}$

by fitting it to a function of the form

${\hat {F}}(x)=\sum _{i=1}^{n}\beta _{i}f_{i}(x)$

where you've already chosen the $f_{i}$ functions and only the $\beta$ 's need to be determined. The number of data points $m$ generally will exceed the number of functions in the model, $n$ , and the functions $f_{i}$ can be anything, not just $x^{i}$ as in the case of polynomial curve fitting.

I don't believe the Ruby and Tcl solutions on the page solve the general case of this problem because they assume a polynomial model. I propose that the task be clarified to stipulate that the inputs are two tables or matrices of numbers, one containing all values of $f_{j}(x_{i})$ and the other containing all values of $F(x_{i})$ with $i$ ranging from 1 to $m$ and $j$ ranging from 1 to $n$ , and $m>n$ . --Sluggo 12:52, 9 August 2009 (UTC)

The thing that worried me was that there wasn't any code in there to determine whether the degree of the polynomial selected was justified. I had to hard-code the depth of polynomial to try in the example code. I think the fitting engine itself doesn't care; you can present any sampled function you want for fitting. (I suppose many of the other more-advanced statistics tasks have this same problem; they require an initial “and magic happens here” to have happened before you can make use of them.) —Donal Fellows 13:47, 9 August 2009 (UTC)

You can always fit any empirical data to a polynomial. This task is about fitting it to functions of a more general form (i.e., a different basis). For example, by selecting the functions

f_{i}

as sinusoids with appropriate frequencies, it should be possible to obtain Fourier coefficients (albeit less efficiently than with a dedicated FFT solver). I don't see how the code in the given solutions could be used to do that. --Sluggo 16:57, 9 August 2009 (UTC)

Well, the only code in the Tcl example that assumes a polynomial is in the example part and not the core solution part; that's using polynomials of up to degree 2 because that's what the page that provided the data used for the example suggested, not out of some kind of endorsement of polynomials. —Donal Fellows 20:51, 9 August 2009 (UTC)

Fair enough. Thank you for the explanation. --Sluggo 22:44, 9 August 2009 (UTC)

Misleading Note in Intro

The note in the introduction is incorrect. So I deleted it. This task is a multiple linear regression problem; the use of OLS indicates that we are dealing with a linear model. This is very different from a polynomial fitting problem which, by definition, is generally non-linear. At best, the multiple regression task is multi-linear and it is most certainly a subset of polynomial fitting problems. The note would be correct if we were talking about a multi-variate polynomial fitting task (which would actually make an excellent task). --Treefall 22:36, 20 August 2010 (UTC)

Python example is not correct???

The Python example solves a different problem, namely fitting a quadratic polynomial in one variable to a set of points in the plane. What is asked for is a way of fitting a _linear_ poynomial in _several_ variables to a set of points in some dimension. I think np.linalg.lstsq() , a function from numpy , is what is needed.

The method with the matrix operations was basically correct, but it was hard to see with it using random data. I substituted the Wikipedia example data so it was more clear that the method works. I also added a np.linalg.lstsq() version, which I understand is preferred. —Sonia (talk) 21:37, 3 April 2015 (UTC)

Task description has too many equations and not enough guidance

Could someone at least try and mitigate the need to at least think you know what multiple regression is, before you can make sense of the page?

The task doesn't ask for an example computation, and the wikipedia page on OLS is impenetrable to the good programmer/lesser mathematician. --Paddy3118 07:13, 2 April 2011 (UTC)

Equations invisible on OS X Chrome and Safari

The main page of this task is one of a number of Rosetta task pages which simply show a set of framed blank panels (not even a fallback image) in lieu of equations, when viewed on OS X with Safari or Chrome (FWIW the MathML on this Talk page is displaying without problem in the same browsers). Is it possible that the markup on some of these pages is making assumptions that are now out of date, or are possibly less platform/browser independent than was hoped at the time of editing ? Hout (talk) 12:46, 15 September 2016 (UTC)

Experimenting with edit preview, I found that visibilty of the equations could be restored (and removed again) by removing (or re-introducing) the single white space between the opening <math> tag and the first printing character of the Math ML code. For the moment I have committed that edit, and can now see the formulae on the OS X (Sierra) browsers. Has that temporary edit suppressed visibility on any other platforms ? If not, I suggest a check for other instances of a math tag followed by an immediate white space character, on other Rosetta Code Task pages. Hout (talk) 13:12, 15 September 2016 (UTC)

Task description now restored to visibility, though excessive font size disparities introduced by doubled <big> tags still need to be be removed Hout (talk) 19:42, 28 September 2016 (UTC)

Font size disparities now normalized by Tikkanz 2016-09-28 Hout (talk) 20:34, 28 September 2016 (UTC)

Some formula elements in the preamble to the Perl 6 contribution remain invisible to most browsers. Two instance of the Greek letter gamma (γ) and the whole of the formula I = h 0 ∧ h 1 ∧ h 2 which follows the word 'Noting'. The server-side graphic used by Chrome, IE/Edge, Safari etc is not displaying. (Only a minority of browsers (using MathML + local fonts, as in the case of Firefox) are displaying these formula elements) Hout (talk) 19:42, 28 September 2016 (UTC)

Examples are not multiple regression

See above for older discussion showing confusion due to proliferation of single-regression examples

Many if not most of the examples on this page were apparently translated blindly from a single incorrect example -- the height/weight examples in particular are all single-axis. Regression of a single X axis is not multiple regression -- see References below. Much of the confusion in this discussion, and many of the suggestions for merging with Polynomial Fitting, stem from that error.

Multiple regression, as mentioned in the description, is where the X input is a N-dimensional array; a matrix of two or more columns, not a list.

History of error

The very first example, Ruby, was incorrect -- it only used a single-axis X input, when the task description clearly said "x is given to you as a two-dimensional array".
The next example added, Tcl, introduced the single-axis height/weight example. Things sort of went all to heck from there, with other languages translating the incorrect example from Tcl.

Examples of multiple regression

Think of multiple regression as "surface fitting" rather than curve fitting. Some examples:

You want to use multiple regression to create a three-dimensional surface so you can estimate implied volatility given time to maturity and delta of an options contract. For pretty pictures of this, see volatility surface graphs. In this case, X is a two-dimensional matrix of {time[i], delta[i]}, and Y is a matching list of implied_volatility[i].
As a possible modification to the existing height/weight examples: You want to use height and body-mass index (BMI) to create a three-dimensional surface so you can estimate weight: X is a two-dimensional matrix {height[i], bmi[i]}, Y is weight[i].

Summary for clarification:

Examples on this page must support a two-dimensional matrix as the X input, as mentioned in the task description.
I've flagged C, Go, and Python as incorrect; that should help seed some multiple regression examples.
The existing height/weight examples could be moved to either the Polynomial regression page if they are performing a polynomial curve fit, or to a Linear regression page if they are just fitting a straight line.

References:

Multiple Regression page on Wikiversity
Multiple Regression description in Wikipedia

--Stevegt (talk) 03:04, 9 September 2023 (UTC)

First, do not top-post, that's just plain rude. I've removed the incorrect markers and reverted the task description. It is very rude to storm in here and unilaterally change a long-standing task description without discussing it first. Remember this is a site for comparing programming languages, and not some kind of "Maths Overflow". If you are going to use terms such as "implied volatility" or "options contract" you need to explain them. If a task needs more precise inputs and expected outputs it should be discussed on the talk page. If this task deserves to be renamed and a new one created that's what should happen, after discussing it and if no-one objects. --Petelomax (talk) 12:14, 10 September 2023 (UTC)

PS If you're going to claim some entries are incorrect, at the very least you should list some which are correct, and if there are none create one, or failing that show some expected/desired output, and if that is beyond you start a discussion on some other maths forum. --Petelomax (talk)

PPS Welcome to rosettacode, and don't fret about any of this, no-one is actually angry with you or anything like that. :-)