When does FORECAST work best?

[S.H.A.D.O.]

Good evening,

I have a column of data of about 2,500 rows.
I am using the FORECAST function on all of the 2,500 rows at present to see how it works!
I know that this number of rows pretty much makes what results FORECAST produces useless.

Does anyone know what a good number of items or rows are to make using FORECAST give realistice and useable results please?

Thanks in advance.

 

[shg]

Forecast does a liner extrapolation, unrelated to the number of data points, just like a linear trendline extended forward. If your data is perfectly linear, it's perfect.

 

[AhoyNC]

The forecast function in Excel uses linear regression. So one question is your data linear?
See if this link on forecasting is of any help to you.

http://highered.mheducation.com/sites/dl/free/0072443901/21708/ste43901_ch03qxd.pdf

 

[S.H.A.D.O.]

Thanks for the replies shg & AhoyNC.

I just thought that because all the numbers were 59 or less that it would give something like an average rather that prediction. That's why I wondered what would be a good number of items to include, something like 50 or 100 etc.

I will have to leave reading the PDF until I have a bit more time so I can take it in.

Thanks again.

 

[AhoyNC]

In the PDF on page 103 there is a table that gives an idea of how much historical data is needed for the different forecasting methods. On page 97 it shows how to use MAD and MSE to measure forecasting accuracy.
Another thing to try is use the forecast function on say the first 50 data points you have and see how well it predicts the next data point.