If we have $n + 1$ points $(x_i,y_i$), then we can use interpolation methods (Lagrange, ...etc) to find a polynomial of degree $n$: $$P_n(x) = a_0 + a_1x + a_2x^2 + \cdots a_nx^n$$

In curve fitting, we search for a polynomial that best fits the points (minimizes the squared error).

But what if we want to use polynomial curve fitting (a.k.a polynomial regression)to find the polynomial of degree $n$: $$Q_n(x) = b_0 + b_1x + b_2x^2 + \cdots b_nx^n$$

After testing this in Excel, I found that the interpolation and the polynomial regression of degree $n$ give the same polynomial.

So my conclusion is that: Using polynomial interpolation through $n+1$ points is equivalent to use polynomial curve fitting of degree $n$.

Am I right?


(I am refering only to polynomial interpolation not piece-wise interpolation)


1 Answer 1


You are right. Note that in both cases, we have $n+1$ unknown coefficients. Finding the interpolation polynomial of degree $n$ means solving a system of $n+1$ linear equations; one corresponding to each point $(x_i,y_i)$. In general, when applying regression, there are many more data points than there are coefficients. The coefficients are then found by solving the least-squares system. However, in your case the polynomial can be fitted exactly, as there are just as much data points as there are coefficients.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .