Ever heard of "Goodness of fit?"
Can you specify exactly what you have forecasted? Is it a single value?
I need to 'formally test' a forecast, I have the actual data and have created a forecast based on a computed beta (finance) and presume I need to use hypothesis testing to formally test the difference between the forecast and the actual.
Does anyone have any advice on what test to use? I am not sure whether a paired t test is appropriate because I am not certain that the two data classify as paired because all the examples I have read describe paired as the same person tested twice. Should I just look at a t test for difference between means or do you have another suggestion?
Thanks for your help.
Thanks yes I studied it a few years ago.... I will have another look. Based on 4 years of historical data we calculated a beta coefficient to make a forecast for the 12 months of 2008. Then we needed to formally test our 12 months of 2008 to the actual 12 months. Goodness of fit still?