Friday, November 21, 2008

Statistics Skills and Software

I'm starting to think that I need to develop some substantial statistics skills. There are a few different reasons for this: I'd like to be able to assess studies I read better, I'd like to be able to do my own studies better (will need to do analysis on some survey results in a few weeks) and I'm getting interested in doing more policy work.

In terms of statistics software, my university provides discounts to purchase Stata and SPSS. A friend of mine doing doctoral work in economics prefers Stata and I am inclined to take her lead on this.

Does anybody reading have any preferences on statistical software or recommendations on how to get a grounding in applied statistical skills?

4 comments:

pakurilecz said...

"I'm starting to think that I need to develop some substantial statistics skills."

good! but have you properly defined the metrics that you will use in your statistical analysis?

'Does anybody reading have any preferences on statistical software or recommendations on how to get a grounding in applied statistical skills?"

first if you don't have a solid grounding in statistics or how to interpret them you might want to get a basic book such as

http://www.amazon.com/Statistics-Dummies-Deborah-Rumsey/dp/0764554239

along with the workbook
http://www.amazon.com/Statistics-Workbook-Dummies-Lifestyles-Paperback/dp/0764584669

always keep in mind what Disraeli said "There are three kinds of lies: lies, damned lies, and statistics."

you need to define your metrics first, what are you measuring so that you can do some meaningful statistical analysis. If your metrics aren't good the analysis is cr*p

for example years ago I worked for an archives that was part of a public library. Once a year the library collected a variety of information. number of patrons, items requested etc. All well and good if you are a library. but as an archives the information collected was not worth a darn. We were only allowed to count a researcher once not all the times we interacted with them during the day. A library patron typically asks a question and goes away. An archives researcher comes in early and stays the day interacting with all the staff who are desk duty during the day. A researcher requests one box and works with it all day. a library patron asks for a book and leaves. both are counted as one item

so my point is make sure what you measure will be meaningful

Russell James, CA said...

Both Stata and SPSS are for hard-core statisticians, people who have a good grounding in their stats education. There are a number of basic stats software programs out there that are open source, too. Look but don't buy something until you know what you need - I agree with Peter on this.

Dee Dee said...

Research design will influence how you collect and analyze statistics, so I'd start with a good book on research design. I liked Douglas Mook's Psychological Research (you could probably get an older edition for cheap) and I think it would be accessible even if you don't have a psychology background.

pakurilecz said...

if you go here you see a number of freeware/shareware statistical software packages

http://www.download.com/1770-2010_4-0.html?query=statistics&searchtype=downloads&filter=platform=Windows&filterName=platform=Windows&tag=ltcol