A*******s
发帖数: 3942 | 1 Does anyone have experience on fitting mixed effect model with more than 500
,000 subjects/clusters, two binormal random effects and 20 fixed effects? I
am trying nlmixed but it has been running over 5 hours. Will Glimmix be
better?
Since the model is of binary outcome I cannot use Hpmixed. :( Any other
suggestions? Thanks in advance. | o****o
发帖数: 8077 | 2 Half million subjects? I doubt if your NLMIXED can finish ever | s*r
发帖数: 2757 | 3 try stochastic gradient descent
haha | z**********i
发帖数: 12276 | 4 今年,我们大的数据,准备在WORK BENCH(UNIX)上试试.
去年,我RUN的NLMIXED,PC上,用了5-7天,给了结果. 后来,用GENMOD了.
500
I
【在 A*******s 的大作中提到】
: Does anyone have experience on fitting mixed effect model with more than 500
: ,000 subjects/clusters, two binormal random effects and 20 fixed effects? I
: am trying nlmixed but it has been running over 5 hours. Will Glimmix be
: better?
: Since the model is of binary outcome I cannot use Hpmixed. :( Any other
: suggestions? Thanks in advance.
| A*******s
发帖数: 3942 | 5 the saddest thing is not that your NLMIXED cannot finish, but rather that
the NLMIXED finished after 20 hours and then you realized you
overparameterized the model... DAMN!!!!! :)
【在 o****o 的大作中提到】
: Half million subjects? I doubt if your NLMIXED can finish ever
|
|
