topic Re: Analysis Question in Statistical Procedures

Analysis Question

OS2Rules — Mon, 30 Nov 2009 14:14:12 GMT

Hi All:

Being a programmer, I'm no great shakes at statistics.

What I have is a measurement for each quarter for a year. I have a value for the first quarter of the following year. (I have YEAR, QUARTER, MEASURE)

I would like to know whether the new quarter is "reasonable" given the values of past year (ie: should I reject that value and investigate).

Remember, I'm a programmer so the simpler the better!

Thanks in advance.

Re: Analysis Question

sbb — Mon, 30 Nov 2009 16:17:29 GMT

PROC COMPARE with suitable options to satisfy your "reasonable" check. Or a DATA step process to compare related consecutive "period" values.

Scott Barry
SBBWorks, Inc.

Re: Analysis Question

plf515 — Sat, 12 Dec 2009 01:34:51 GMT

OK, so you've got something like this

2008 1 10
2008 2 12
2008 3 11
2008 4 10
2009 1 9
2009 2 10
2009 3 11
2009 4 10
2010 1 xxx

then you find that xxx = 12. Is 12 "reasonable".

Is that your question?

If so, may I ask how many years of data you have, and whether there is reason to expect either a long term trend or seasonal variation or both?

Peter
http://www.statisticalanalysisconsulting.com/

Re: Analysis Question

OS2Rules — Wed, 16 Dec 2009 18:57:00 GMT

That is exactly what I am looking for.

I only have 2 years of good data (8 quarters) which I know is a problem (not enough data!)

I would like to know that when xxx=12 it is "ok" and when xxx=345 is it not (or whatever) based on the previous set of data for that measure.

Re: Analysis Question

plf515 — Thu, 17 Dec 2009 02:34:57 GMT

Sorry, but I don't think there is a good statistical answer for this. You can get a distribution of quarter to quarter change, but there N will = 7, and that's not much to go on; esp. if there are seasonal patterns.

Instead, I think you have to substantively define what would be an unreasonable change.