Message-ID: <loom.20111021T163318-491@post.gmane.org>
Date: 2011-10-21T14:35:08Z
From: Ben Bolker
Subject: glm-poisson fitting 400.000 records
D_Tomas <tomasmeca <at> hotmail.com> writes:
>
> Hi,
>
> I am trying to fi a glm-poisson model to 400.000 records. I have tried biglm
> and glmulti but i have problems... can it really be the case that 400.000
> are too many records???
>
> I am thinking of using random samples of my dataset.....
>
"I have problems" isn't enough for us to diagnose. I tried
this trivial example in base R:
> d <- data.frame(x=runif(4e5),y=rpois(4e5,5))
> system.time(glm(y~x,family=poisson,data=d,trace=TRUE))
Deviance = 438614.6 Iterations - 1
Deviance = 417968.2 Iterations - 2
Deviance = 417921.2 Iterations - 3
Deviance = 417921.2 Iterations - 4
user system elapsed
5.444 12.952 18.429
Can you give us a hint about what went wrong??