A statistical model can be considered reliable mainly when it is capable to explain the variance of the data. A robust model should also consider validity in addition to reliability. This condition implies to include an analysis about three crucial aspects: convergent, discriminant and nomologic validity. These criterias have different ways to be proven in order to ensure your model fits the data and includes the adecuate variables. So references from the multivariate statistics are required to.
Curve prediction is a very complex subject and there are a lot of methods to follow. But before analysing the data it is quite important to filter and normalize. Once you choose the appropriate method, you have to consider all the collateral proofs required. Test and retest for evaluating the eficiency of the model.
I specially recommend you regression, because of its wide employment in marketing analysis. I would really like you to be more explicit about the series, variables and points of view in your research to be able to give you a more structured answer.
HOPE TO BE USEFUL
Hello
I think the best method is the regression. Even the most simple linear function. After its, you should use a different method of verification of the forecast, for example formula by Teill
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics