Improving the code to change the dimension of the variable.(about the autocorrelation and autocorr)

I have to write the code like the question said,and here is my code

 N=16;
 fD_Ts=0.001;
 for n = 1:1:100
 theta_n(n)=0+(2*pi)*rand(1,1);
 phi_n(n)=0+(2*pi)*rand(1,1);
 a_n=( (2*pi*n+theta_n)./N )

    for m=1:1:1000
        h_c=(1/N)*cumsum(cos(2*pi*m*fD_Ts*cos(a_n))+phi_n);
        h_s=(1/N)*cumsum(sin(2*pi*m*fD_Ts*cos(a_n))+phi_n);
        h_total=h_c+1i*h_s;
 answer=autocorr(h_total);       
    end
 end

Before looking the code,we know if we want to plot the (n,autocorrelation),n and autocorrelation have to be the same dimension,that is,1 by 100 matrix.However,i found that the dimension of answer is just 1*21 double,why?how should i improve my code to let the dimension of answer become from 1*21 to 1*100 ?

I don't know how to write the autocorrelation like the formula,if writing that formula is the only way,can anyone teach me how to write that code?Thanks a lot