hjk_nll_wrapper: A wrapper for calculating the negative log-likehood in...
In MichaelHoltonPrice/yada: Yet Another Demographic Analysis package

This wrapper function for doing the multivariate negative log-likelihood calculation accomplishes three things. First, if the negative log-likelihood evaluates to NA it replaces the value with Inf. Second, it rescales the normalized, input parameter vector (param) to account for the offset (th_y_bar0) and scale (th_y_bar_scale); in particular, the rescaling is param = th_y_bar0 + param*th_y_bar_scale. Third (optionally), it saves to file progress information on the optimization (notably, the best fit found so far). This progress information can be used resume interupted optimizations.

hjk_nll_wrapper(
  param,
  th_y_bar0,
  th_y_bar_scale,
  calc_data,
  tf_cat_vect,
  save_file = NA
)

`param`	The normalized parameter vector
`th_y_bar0`	The offset to use for the normalized parameter vector
`th_y_bar_scale`	The scaling to use for the normalized parameter vector
`calc_data`	The calculation data that support rapid calculation of the negative log-likelihood
`tf_cat_vect`	The transform category vector so that the optimization can be uncsontrained
`save_file`	(default: NA, none used) A save file to capture optimization progress