Framework for the Analysis of Factorial Designs

Pre-analysis, which characterises the experiment and determines the appropriate method to be followed. 

Calculating the values, using appropriate formulae compute the parameter estimates, the normal probability plot of the parameters, and if appropriate, the SSRi and F-value for each of the parameters. 

Model reduction, which seeks to remove all uimecessary parameters from the model. Three main methods can be used  

Residual analysis, using the reduced model, the resulting residuals should be analysed to determine whether the assumptions underlying regression have been satisfied. If the residuals suggest problems, then further experiments may need to be performed to resolve the issue. 

Conclusions, which seek to answer based on the reduced model whatever questions the objectives of the exercise had, for example, determining the optimal operating point. 

---