Hi all,
In using the lme4 package to construct multi-level models, I notice that
for large datasets (~300000 data points, >100 groups, possibly cross/nested
effects) model fitting can be very slow (>2 hours). I have even larger
data sets that I'd like to model, but hesitate due to concerns about run
time.
To that end, is it possible to parallelize/multithread lme4?
Alternatively, can SparkR be configured to run lme4? Finally, if neither
is possible, what can be done to speed up lme4?
Thanks,
Sue