I have an observed data set $O$ and a synthetic model $S(\theta)$ which attempts to describe it.

By fixing $\theta$ to different values I can generate $M$ synthetic realizations of the model:

$$S_k = S(\theta_k),\, k=1..M$$

I want to find the $S_k$ that best matches my observed data set $O$.

I use a genetic algorithm to find the maximum of a likelihood equation $L$ between the $S_k$ and $O$. The $S_k$ with the maximum likelihood value is:

$$S_{max} = S(\theta_{max}) = max[L(S_k, O), \, k=1..M]$$

$\theta_{max}$ is thus the value of $\theta$ that I associate with my observed data set $O$.

To obtain the uncertainty of $\theta_{max}$, I repeat the above process $n$ times on random samplings with replacement of the true data set $O$, ie: I apply a bootstrap with replacement process.

The bootstrap returns $\theta_i, \,i=1..n$ values of the model variable.

I obtain the uncertainty on $\theta_{max}$ by calculating the standard deviation of the $\theta_i$ values:

$$\theta_{max} \pm std[\theta_i, \,i=1..n]$$

But I can also use the $\theta_i, \,i=1..n$ values to obtain a mean:

$$\theta_m = mean[\theta_i, \,i=1..n]$$

My question: what is the relation between this $\theta_m$ and the $\theta_{max}$ value obtained via the maximum likelihood?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.