Why do we divide by n-1 when computing a sample variance? Here is a step-by-step proof that dividing by n leads to a biased estimate, whereas dividing by n-1 leads to an unbiased estimate.
In SEM and factor analysis, how does one derive a discrepancy function from a likelihood function? Here is a derivation of the ML discrepancy function from likelihoods, with or without a mean structure in the model, and using either raw data or a covariance matrix as input.