A logical argument to calculate residuals or not would speed up model fitting, especially useful for lfo_cv runs. Would also lead to smaller model objects in terms of memory.
Could also reduce the parameters being monitored during a lfo_cv refit to also speed up post-processing (only really need mus and family-specific parameters for calculating log-likelihoods; or ypred for other types of scores).
Finally, could avoid adding estimates to the mgcv model and avoid returning the mgcv model to also improve speed in lfo_cv runs
A logical argument to calculate residuals or not would speed up model fitting, especially useful for lfo_cv runs. Would also lead to smaller model objects in terms of memory.
Could also reduce the parameters being monitored during a lfo_cv refit to also speed up post-processing (only really need mus and family-specific parameters for calculating log-likelihoods; or ypred for other types of scores).
Finally, could avoid adding estimates to the mgcv model and avoid returning the mgcv model to also improve speed in lfo_cv runs