Dear all
I've had a painful afternoon trying to get R to run an LME analysis using the same scripts/processes that I've successfully employed previously. No joy this time. When examining the structure of the imported data [str(dataset)], I discovered that R has changed all numerical variables to factors, and all factors to integers. Any idea why this is happening, and how to overcome?
I've used some pretty advanced swear words this afternoon.
All numerical variables to factors? And all factors to integers? This is a strange behavior which I did not confront so far. Could you share some part of csv file and your import command?
Sympathies for the frustration! Were you using read.csv or read.table with default options? That can work out okay with some input files and not others.
Instead, if you specify how you want each column to be converted (use option colClasses, see ?read.table for details) then you should get exactly what you want.
75
It's like North Korea, isn't it? A supposed paradise that it's actually near impossible to get things into our out of!
Dear all - thanks for the collective sympathies. Certainly helped to raise a smile which was most definitely needed!
A colleague has pointed out an issue in the source file which seems to have underpinned the current problems. The .csv was generated by exporting a dataset from SPSS. SPSS has (for reasons of its own imagining) inserted the phrase '#NULL!' into cells with missing data - removing the '#NULL!' and leaving blank now permits R to recognise these columns as numerical. Stupid error - but at least I've learned something.
The second issue RE conversion of factors->integers appears to be a problem with the coding used for our clinical trials. We currently use binary code where 0=No and 1=Yes - R has interpreted this as integers, although as this affects relatively few variables manual conversion (e.g. as.factor(variable)) isn't so galling a task.
@Julia - the file was imported as read.csv(' data.csv', strip.white=TRUE). I note that you're a fellow JCU minion, so will buy coffee if you can stomach an R discussion!
Thanks again
J
Dear Joseph,
the standard read functions of R (including read.csv) have also the argument "na.strings" and na.strings = "#NULL!" might also have worked for you.
An interesting alternative is package readr of Hadley Wickham that has function read_csv (not automatically converting strings to factors). It's on CRAN, but see also https://github.com/hadley/readr.
Best
Matthias
Thanks all @Maximilian - that package sounds particularly useful. I'm the only one in the unit using R - everyone else is an SPSS person!
check function arguments
as.is
and
colClasses
in read.table when using R. You can explicitely define the class of each variable when importing the csv data set.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling