Minimization process¶
Least squares¶
Experimental curves are fitted with models using the least-squares method. The method of least squares is a standard approach in regression analysis and is based on solving the following problem for the coefficients \(x\):
(111)¶\[\underset{x}{\mathrm{min}} \| F(x,xdata)−ydata \|_2^2 = \underset{x}{\mathrm{min}} \sum_{\substack{i}} (F(x,xdata_\text{i})−ydata_\text{i})^2\]
Here, \(xdata\) are the input data and \(ydata\) is the observed output.
Fischer-Cripps A.C. wrote a detailed section about the non-linear least squares fitting in his book “Nanoindentation” (pp. 253-256) [1].
Residuals¶
The residuals from a fitted model are the differences between the response data and the corresponding predictions computed using the regression function. For the ith observation in the dataset, the residual is defined as:
(112)¶\[r_\text{i} = y_\text{i} − \hat{y}_\text{i}\]
Here, \(y_\text{i}\) is the ith response value and \(\hat{y}_\text{i}\) is the ith predicted response value.
If the model fits the data correctly, the residuals approximate the random errors.