docs map_estimate about bayestestr HOT 8 CLOSED

Thanks for pointing that out! You're right, I think this is a bad design choice. Better to stick with the return in the form of a dataframe containing the two values. Also, I've added more info about them. Will be in upcoming PR

from bayestestr.

After thinking a bit more I've decided to re-set the inclusion of the density value to optional as I think most users won't expect nor need it. I tried to document it better in the documentation.

from bayestestr.

I am still hesitating with this one... currently MAP_estimate always returns a dataframe with one or two column). However, users might expect a single value instead of a dataframe. The density value of the MAP (the possible second column) was initially added because it is used in p_map and because it might be worth something (#59). At the same time, it can be easily retrieved using density_at, although it is computationally not very efficient (i.e. the density function would be estimated twice).

Nevertheless, maybe it's better to harmonize it with other functions and return a single value (in case a single posterior is passed) instead of always a dataframe...

from bayestestr.

Well, you still have attributes (to save the density value), and you can add a print()-method. What is printed (many information) and what is returned for further computation (a numeric value, with attrbiute) can be something different.

from bayestestr.

storing the density value as an attribute solves all problems. Thanks for that :)

from bayestestr.

Well I am kind of disappointed to find out that numeric cannot really take attributes... 😅 More precisely, adding attributes to a numeric make the tests fail. If I create a new class `c("MAP", "numeric"), the tests also fail... See #67

Returning a list seems a bit as going back to the dataframe behaviour 😕

from bayestestr.

You can actually check for values with attributes in tests, I have done this somewhere in on of my / our packages... I must check if I can find it...

from bayestestr.

I think this has been addressed

from bayestestr.