MATLAB Answers


How to use polyfit to get the difference out of two functions?

I have two datasets (X,Y). For every value of Y that corresponds to a value of X.
I am thinking of applying a polyfit command to get the best curve fitting for the data, and as a result I will come up with a function (slope+intercept).
How do I find the difference between those two functions? What command do I apply to subtract Function A from Function B?


One of the functions is the one implied by polyfit(x,y,1) . But what is the other function?
A collection of points is not considered a function except in mathematical theory (in which case, the collection of ordered values is considered to define the function, and formulae are just shorthand for the collection of values.)
@Walter. The datasets are curves. If I plot them in excel and apply a polynomial curve fitting of high degree (e.g. 6) then I get the best line fit and a function. I need to get the same function in Matlab. And then subtract one function from another since the datasets are very close each other, to find their difference. How do I do that?

Sign in to comment.

3 Answers

Answer by Walter Roberson
on 21 Apr 2019
 Accepted Answer

P1 = polyfit((1:length(X)).', X(:), 6);
Likewise for Y producing P2
Pd = polyval(P1, 1:length(X)) - polyval(P2, 1:length(X))


%after the readtable
mask = isnan(x1{:,:}) | isnan(y1{:,:});
x1(mask,:) = [];
y1(mask,:) = [];
mask = isnan(x2{:,:}) | isnan(y2{:,:});
x2(mask,:) = [];
y2(mask,:) = [];
I get the below error
P1 =
0 0 0 0 0 0 0
Warning: Polynomial is not unique; degree >= number of data points.
> In polyfit (line 74)
In curvefitdif (line 19)
P2 =
0 0 0 0 0 0 0
Pd =
0×1 empty double column vector
P11 =
0×1 empty double column vector
P22 =
0×1 empty double column vector
I also get the error
Error using isnan
Too many input arguments.
Error in see (line 38)
mask = isnan(x1{:,:}) | isnan(y1{:,:});

Sign in to comment.

Answer by John D'Errico
on 21 Apr 2019

Polynomials are linear in the coefficients. So the difference of two polynomials is obtained by just subtracting the coefficients. If they are different order polynomials, just pad the low order polynomial with zeros as necessary to correspond to the higher order coefficients that are effectively zero.


Yours John was also good but I couldn't make the fplots work correctly. I was getting some results, but they were almost linearly, which in my case the data are curves.
If pdiff gives a result of NaN, then one or both of the polynomials was already garbage. LOOK at the cofficients that you got from those two fits.
NaN can only ever result as a subtraction by things like inf - inf, or anything plus a NaN. So the problem lies in your original fits.
Actually, the problem lies where I use readtable and import the columns from excel.
They are imported as NaNs

Sign in to comment.

Answer by Stelios Fanourakis on 21 Apr 2019

Actually, both datasets are high degree polynomials e.g 6. They are almost identical and need to find their difference. How do I do that in Matlab?

  1 Comment

Did I not say just how to do that in my answer?

Sign in to comment.