Since stock prices are the time series data, you might think the time series models are the good models for them., such as ARIMA model, GARCH model, etc. It seems really good, but the data can not satisfy with the assumptions of the models. We need to back the origin of model building.
The regression analysis is to seek finding the linear relationship of the dependent variable and the independent variables. For the time series data, the characteristic is “time”! There are rarely researches are willing to discuss this kind of model where there is a time variable as an independent variable. Researchers considered this kind of model does not have any sense for researches. This is really a pity！
So, I back to the origin of linear regression model and rethink that:
- Do not put all data to fit model, but choose the optimal sample size to fit.
- From the start of data period, given five samples to fit. Add additional one sample to fit model. Compare which one has the higher precision, and the samples become a group.
Now, I can share you my examples.
Big Data analysis and simulator 