major update to the smoothic() function to include
different families of distributions
addition of the “smooth generalized normal distribution”, where an additional shape parameter is estimated relating to the kurtosis of the error distribution (shape parameter can also be fixed at a user-supplied value)
new option to use nlm() for optimization
(optimizer = "nlm") or to use the manually coded
Newton-Raphson method (optimizer = "manual")
addition of the Laplace distribution, which corresponds to robust regression where the errors are heavy-tailed
new dataset bostonhouseprice2, which is a corrected
version of the original bostonhouseprice data
new dataset diabetes
initial release
two datasets bostonhouseprice and
sniffer
automatic variable selection using the smoothic
function
can choose between distributional regression (multi-parameter)
with model = "mpr" and location-only regression (single
parameter) with model = "spr"