Hello Shuiyin,
Yes, it is usually better to project the data. This is especially
the case if you are calculating any of the endemism indices as the
ranges are analysed in units of grid cells, for which an equal area
projection will ensure they all occupy the same amount of area on
the earth's surface. (Although in saying that, it does not correct
for cases such as coastal cells which are part land and part water
and thus for many taxa will be an overestimate).
A useful projection to try in the first instance is the Mollweide,
but there will be others.
https://epsg.io/54009
You are also correct that you will also need to project any
shapefile you want to display in Biodiverse so its coordinates match
your points.
In terms of what cell size to use, it really depends on the spatial
density of your data. As a general suggestion, if the extent is
global then 100km will probably be sufficient to see many of the
patterns. However, if you have good data then why not go finer? If
you need smoother patterns to aid interpretation then you can try
the moving window spatial conditions like sp_circle() and
sp_block().
https://github.com/shawnlaffan/biodiverse/wiki/SpatialConditions
One additional consideration for cell size is the resultant size of
the data sets and how much memory you have available on your
computer. For comparison, the data sets in Mishler et al. (2020,
https://doi.org/10.1111/jse.12590 ) used ~15GB for display. That
was for ~10,000 cells with a tree containing ~20,000 terminals, and
multiple display tabs open (each of which needs its own memory
allocation). If you do not have a tree then less memory is needed.
Regards,
Shawn.