If we have \(k\) parameters to estimate, we can solve this if we have \(k\) equations.
We generate these
First, we link each first \(k\) moments to functions of the parameters.
Then we replace the momenets with sample estimates.
The moments of this population distribution are:
\(\mu_i =E[X^i]=g_i(\theta_1,...,\theta_k)\)
We have a sample.
\(X=[X_1,...,X_n]\)
We now define the method of moments estimator
\(\hat \mu_i=\dfrac{1}{n}\sum_{j=1}^nx_j^i\)