n = 0:15;
x = 0.84.^n;
y = circshift(x,5);
[c,lags] = xcorr(x,y);
stem(lags,c)
xlabel('lag')
ylabel('r')
or more simply,
plot(lags, c, '-o')
This shows a maximum correlation at lag -5 which makes sense since we circularly shifted the data by 5 units.