How to decide which window function (Hamming,Gaussian) is best fit for our data?

I have generated a white gaussian noise with fixed standard deviation and mean=0 as; load B_oct1 % load fast fading signal of length 3661 Sfixed=0.1445; % Sigma Measured mu=0; n2 = normrnd(mu,Sfixed,[length(y) 1]); then I took Autocorrelation of n2 MAXLAG=21; [Rx ,lags]=xcorr_unb(D(n2, [], MAXLAG, 1, 1); % Calculate autocorrelation cuttoff=[0.01 0.07]; % 2 element vector as for bandpass fast fading is bandpass signal order=20 h=fir1(order,cutoff,gausswin(21)); % filter coefficients I choose here Gaussian window as fast fading signal seems to have Gaussian distribution.

I want a best fit of filter coefficients with autocorrelation of the fast fading? Which window to use and how we can find a better fit?