In Estimation method, choose between the least squares estimation (LSE) and maximum likelihood estimation (MLE) methods. The methods produce equivalent estimates of the coefficients in the saturated model, when the number of parameters equals the number of data points.
Often, the differences between the LSE and MLE results are minor, and the methods can be used interchangeably. You might want to use both methods and determine whether the results confirm each other. If the results differ, you might want to determine why. For example, MLE assumes that the original data are from a normal distribution. If your data might not be normally distributed, LSE may provide better estimates. Also, LSE cannot calculate results for data that contain a standard deviation equal to zero. MLE might provide estimates, depending on the model.
You can use LSE and MLE together because LSE provides better p-values, while MLE provides more precise coefficients1
. To use this approach, follow these steps:
- Use the p-values from the LSE to determine which terms are statistically significant.
- Fit the model again, excluding nonsignificant terms to identify the appropriate reduced model.
- Use MLE to estimate the final coefficients of the model and to determine the fits and the residuals.