Forecasting :( need help

[a_masry]

Hello all of you ,
I need forums talking widely and in details .. about Forecasting studies and problems .. also Statistics
If this GREAT forum contain a part for that .. tell me where it's ?
if not .. plz tell me other resources ..
thnx in advance

[Wes Bucey]

Originally Posted by a_masry

Hello all of you ,
I need forums talking widely and in details .. about Forecasting studies and problems .. also Statistics
If this GREAT forum contain a part for that .. tell me where it's ?
if not .. plz tell me other resources ..
thnx in advance

Welcome to the Cove!

Tell us about your situation. Forecasting is a VERY big field. Let's start by telling us what you want to forecast: sales? Return on investment? Market penetration? Design of Experiments? etc.?

Similarly, Statistics is also a big field. It is much easier for us to answer specific questions than to try to give you an entire education in the field.

You are entering the weekend here in the USA and many of our Covers take a break from the Cove, so answers may be scarce until Monday.

For a thorough background in statistics and a wonderful free handbook on even the most intricate aspect of statistics, take a look at
http://www.itl.nist.gov/div898/handbook/index.htm

[a_masry]

thnx for your reply..
yea I know , it's veryyyy wide field ..
and i have some specific questions and may have more soon,, But I did find a part here in this forum for this WIDE field .. so I thought that It is related to another place for discussion not here .. just to be more organized..
But as I can ask here ,, well
I have data now (numbers) how can I know it's random or not ??
and also how can i test homegenity ?(i.e: when can i say this data is homegenious? )

[Tim Folkerts]

For a set of data to be considered random, there are (at least) two conditions that should be met.

1) the overall distribution of numbers should follow some distribution - perhaps normal, Poisson, uniform, etc. You can do a rough check by by looking at a histogram and seeing if it looks like the distribution you expect. There are, of course, also more mathematical tests to see how close it is to a given distribution.
2) there should be no way to predict the next result given the previous results. This is a little harder to judge by eye, but Minitab has various functions to study this under the "Time Series" submenu within the "stats" menu.

If you post the raw data, I (or someone else) might be able to look at it a bit. You could always remove any labels so we don't know just what kind of data it is precisely.

Tim F

[Steve Prevette]

When it comes to forecasting, throw out your traditional moving averages and regressions. Understand the system, understand the system's variation in its data, and use SPC.

http://www.hanford.gov/safety/vpp/trend.htm