This thesis uses parameter clustering and restrictions on parameter magnitudes to achieve two different goals:
- accurate forecasting and
- valid inference in a data-rich environment.