Skip to content
Prev 17949 / 20628 Next

Large mixed & crossed-effect model looking at educational spending on crime rates with error messages

Despite the extension, it's not a CSV file, it's RData which R does
actually zip depending on the options you passed -- readRDS() works.

@Ben: there's a null byte in there, which is why your system
(rightfully) though it was a binary data file.

Ben's and Thierry's comments cover all the other issues, I think.

My advice would to be ignore the random effects bit for a little while
and think about how you would structure your fixed effects for a plain
lm() so that you can answer the inferential questions you care about.
After that, I would come back to the random effects. State, county and
place all seem like fine blocking variables, but nesting race within
place really doesn't make sense for you current fixed-effects structure.?

Looking at the dataframe, I also have the strong suspicion that it's not
quite structured how you think it is. Right now, per_race is a single
column, which means you have each observation repeated five times for
each race. (This ties into Ben's second comment.)

Phillip
On 28/09/2019 04:14, Ben Bolker wrote: