nonuniform sampling rate and lsqnonlin

Hello all, Imagine we have a signal, S_t(theta), of which we have a noisy observation, y_t, with unequal sampling rates. Assume noise is gaussian with constant variance. Now, my sampling rate is unequal as i said: I have 120 ( length(y_t)=120 ) data points, first 60 points are sampled at 1 sec intervals, last 60 points are sampled at dt=4sec. Now, in the least square fitting process of lsqnonlin, how does the variance of the noise change according this sampling rate? Basically, what do I minimize here:

fun  = ( abs(y_t_est - y_t) ).*sqrt(dt);
or  

fun  = ( abs(y_t_est - y_t) ).*sqrt(1/dt);

Note that fun is what s being passed to lsqnonlin for minimization. and y_t_est is my estimation of y_t based on whatever model I have assumed. Thanks in advance.

[J J]

and note that lsqnonlin on its own squares whatever fun is.