Experimental error and curve fitting

Hi all

Context: I have fit some data to a function using excel solver to minimise the error squares (sum of squares method).

Question: How I do I determine an appropriate error given experimental error as well as goodness of fit?

I know i can calculate an error due to the goodness of fit directly from the sum of squares:

σ² = ∑(y_experimental - f(x))/(N - 1)

so: y_calculated = f(x) ± σ

But if my experimental data has an error (for example, reading temperature off a thermometer with an error of ±0.05, such that each value of y_experimental has an error of 0.05) how do I account for this error in σ (if necessary)? I will be using these values in further calculations, and know how to propagate errors in such calculations, but I first need an error associated with my function that ideally accounts for experimental error as well as goodness of fit.