There are several scenarios for which you might enter values for estimated coefficients. For example, you may wish to give starting estimates so that the algorithm converges to a solution, or you may wish to validate a model with an independent sample.
- Convergence: The maximum likelihood solution may not converge if the starting estimates are not in the neighborhood of the true solution. If the algorithm does not converge to a solution, you can specify what you think are good starting values for parameter estimates in Starting estimates for
algorithm in the Options subdialog box.
- Validation: You may also wish to validate the model with an independent sample. This is done by splitting the data into two subsets. Use the first set to estimate and store the coefficients. If you enter these estimates in Estimates for
validation model in the Options subdialog box, Minitab will use these values as the parameter estimates rather than calculating new parameter estimates. Then, you can assess the model fit for the independent sample.
In both cases, enter a column with the first entry being the constant estimate, and the remaining entries corresponding to the model terms in the order in which they appear in the Model box or the output.
