Analysis of 2**(K-p) and Screening Designs

Analyzes balanced and unbalanced 2**(K-p) designs, Plackett-Burman screening, or other 2-level factorial designs with or without blocking, and with or without center points. The program will compute a complete analysis, estimate factor effects and regression coefficients, produce a large number of additional diagnostic results and graphs to help assess the goodness of fit for the model and model adequacy, etc., and compute various residual statistics for each run (observation).

 Note: You can also use the General Linear Models facilities to analyze unbalanced and incomplete designs of any complexity.

Element Name Description
General
Detail of computed results reported Detail of reported results; if Minimal detail is requested, the program will compute factor effects and regression coefficients, the overall ANOVA table, and the Pareto chart of effects. At the Comprehensive Level of detail, the program will also report normal and half-normal probability plots of effects, square plots for the predicted response, and report and plot means. At the All results Level of detail, the program will also report cube plots for the predicted response, and surface plots for the fitted model (for designs with fewer than 7 factors). If residual analyses are requested, different residual statistics and graphs are also reported at the Comprehensive and All results Level of detail.
Blocking Indicate whether the design is blocked. If the experiment involves blocking, the categorical predictor variable that contains the blocking information (i.e., the codes to indicate to which block each experimental run belongs) must be the first variable in the list of categorical predictors (this is the default arrangement of blocking and experimental factors if the design was created via the standard Experimental Design tools).
Model Determines which terms will be included in the model. All ANOVA results, effect estimates, predicted and residual values, etc. will be computed based on this model. For screening designs with many (e.g., over 100) factors, be sure to specify the No interactions (main effects only) model.
ANOVA error term Choose between two error terms for the analyses; the selected error term will be used in all tests for statistical significance and in the computation of standard errors. Note that Pure error will be estimated from duplicated (identical) experimental runs, so if no such runs are available, this option is not applicable.
p, for confidence limits Specifies a probability value for establishing confidence intervals for parameter estimates, predicted responses, etc.
p, for highlighting Specifies a probability (alpha) value that will be used for highlighting significant results in various results spreadsheets.
Generates data source, if N for input less than Generates a data source for further analyses with other Data Miner nodes if the input data source has fewer than k observations, as specified in this edit field; note that parameter k (number of observations) will be evaluated against the number of observations in the input data source, not the number of valid or selected observations.
Design Properties
Displays design Displays a spreadsheet showing the unique runs (those with unique combinations of factor settings) in the experiment. In addition, for each unique run, STATISTICA computes the mean, standard deviation, and standard error of the mean (if there is more than one run for the respective unique combination of factor settings).
Correlations of effects Choose this option to compute a Correlation matrix of all effects in the current design.
Alias matrix Computes and reports the confounding of main effects and two-way interactions via a correlation matrix of the columns of the design. In this matrix, main effects that were created as aliases of two-way interactions will show a correlation of 1.0 with those interactions.
Blocking generator Creates and reports a spreadsheet containing the generators (aliases) of the blocking factors; only applicable to blocked standard designs.
Residual Analysis
Residual analysis Creates predicted and residual statistics for each run, and compute various results spreadsheets and graphs for these values (depending on the Level of detail selected for the reported results).
Creates statistics If the Extended residual statistics option is selected, the results spreadsheet will report for each run (observation) various residual statistics (in addition to the simple predicted and residual values), such as Mahalanobis distances, deleted residuals, etc. .
Plots residuals Choose which type of residual values to use in plots; if the Plot Raw residuals option is selected, residual plots will use the raw residuals for each dependent variable for each case or run as the values to be plotted. If the Studentized deleted residuals option is selected, studentized deleted residuals will be plotted.
Box-Cox Transformation
Box Cox transformation Creates (find) an optimal lambda value for the Box-Cox transformation; based on the computed value of lambda, you can determine the transformation from the family of power transformations (Box-Cox transformations) that minimizes the error variability (the unpredicted variation) in the dependent variable for the current model.
Max number of iterations Specifies the maximum number of iterations for the iterative search for the best value of lambda.
Minimum value of lambda You can specify the range (minimum, maximum) of values from which to search for the maximum likelihood estimate of lambda.
Maximum value of lambda You can specify the range (minimum, maximum) of values from which to search for the maximum likelihood estimate of lambda.
Delta for convergence, 1E- Specifies the negative exponent for a base-10 constant Delta (delta= 10^-sdelta); the default value is 5 (.00001). This value is used as the target difference in successive estimates of Lambda that will terminate the search.