Size Limitation

[fmgoto]

Are there any limitations to the number of observation? Can I read microdata (hundreds millions)? Thanks.

[EViews Gareth]

15m observations is the limit. Although if you have anything close to that many, you'll probably run out of memory quite quickly.

[startz]

Are there any plans to do a 64-bit recompile?

[startz]

I regularly use EViews for hundreds of thousands of observations, but not hundreds of millions. Just out of curiousity fmgoto, what sort of data are you using?

[fmgoto]

Thanks.

(I'm using household microdata. SPSS or SAS handle those much better, for sure, but it would be wonderful if I could analyze these with Eviews!)

[startz]

I also sometimes use household micro data, but my files aren't quite that large. One thing that sometimes helps is to start with the data in Stata or SAS, dump the variables needed into a separate file, and then read that file into EViews. EViews reads Stata and SAS files really well.

[fmgoto]

I was writing a post at "Data Manipulation " forum, but I think it would be better be discussed here, to avoid losing the thread. Gareth, feel free to move it.

First, I apologize, but I am working with about 20-50 million observations and a dozen fields/series. It is not that gigantic. Given the 15m constraint, I will work with a random sample. That might do the job.

Second, I wonder if there is a way I could import a database (a CSV file) into an Eviews database without reading it (the CSV) into a workfile. Does this 15m constraint apply to .db? files too? Does that 1000Mb restriction on memory size apply to files (.wf1 and .db?) too or just cache?

What is this Eviews Illustrated? I need one of those! May I suggest a more complete treatment on Databases for the next User's Guide? I am a huge fan, but seems not good enough for databases.

SAS/SPSS/Stata would be an option, but people love Eviews around here! It would be a lot easier to work with Eviews, if possible.

Thanks, Fabio

[EViews Chris]

Data from a CSV file does need to go through a workfile before going into anything else.

You can read in a subset of columns (or rows) when reading a csv file so you could break the csv file up into multiple EViews workfiles if you wanted.

I'm not sure this helps you much as far as actually doing analysis on the complete data set though.

By the way, memory constraints inside EViews are a bit complicated. On a 64-bit version of Windows you can probably get a workfile up to around 2 to 3 gigabytes (which is something like 400 million total data points), but some operations inside EViews require working space, so you can get into problems as the workfile approaches 32-bit memory limts. The limit of 15m observations per series is a bit arbitrary, but is chosen to try to minimize people getting into too much trouble.

The solution, as you mentioned, is a 64 bit build of EViews. We are certainly considering this, but I wouldn't expect anything in the near future.

[fmgoto]

Thank You.