Second, you can centre and scale the fit to your data (if necessary) by asking polyfit for two extra outputs, and then passing them to polyval:
[p,S,mu] = polyfit(x,y,n);
[y,delta] = polyval(p,x,S,mu);
See the documentation for a full description of the function options.
EDIT — You can use The File Exchange polyparci (link) function with polyfit to estimate the parameter confidence limits. These are important because any parameter with confidence limits that include zero (that is, have opposite signs), are not needed in the polynomial fit. A model with all parameters having significant (same-sign) confidence intervals (with polyfit polynomial models, different ‘models’ are defined by different polynomial degrees), is the likely the ‘best’ model in this context.
This information will help you refine your model.
