Curve fit using lsq function or something else ?

I have 3 sets of data, that be set-1,set-2,set-3. I have to model set-1 using set-2 and set-3, such that set-1= a* set2 + (1-a)* [damped harmonic convolution-Q] of set-3; There are 2 variables a and Q. I wrote the code for the convolution. I want to know how can I minimise the function effectively using Matlab. Currently I generate random values for a and Q in a range and look for the minimum, however this is very inefficient. Is there a better way ?