Technical Notes - ByGroup Analysis with Variance Estimation and Precision

It is possible to perform by group Variance Estimation and Precision analyses (using STATISTICA's ByGroup Analysis module). When specifying a Variance Estimation and Precision by group analysis, follow these rules to manage the (default) design for the analyses:

  1. If the input Spreadsheet does not contain a previously-saved default design, then

    1. You must specify variables, i.e., the dependent variables, categorical predictors, and/or continuous predictors. As with all Variance Estimation and Precision analyses, you must specify at least one categorical or continuous predictor variable; however, the selection of a dependent variable is optional.

    2. The program will always (for each by group) generate a default design from the specified structure or coding (of categorical predictors) found in data file (e.g., nested effects will be random, crossed effects will be fixed). See Determining the Default Design for more information on how this is done.

    3. A design other than the default design can be specified by typing the respective design into the Design and Random Effects edit controls on the ByGroup Analysis Specifications - Variance Estimation and Precision Advanced tab. Use the same syntax as that which is recorded when creating macros from the Variance Estimation and Precision interactive analysis.

      1. For example, as the design you might specify: GROUP + GENDER + TIME + PAID + GROUP*GENDER + GROUP*TIME + GROUP*PAID + GENDER*TIME + GENDER*PAID + TIME*PAID + GROUP*GENDER*TIME + GROUP*GENDER*PAID + GROUP*TIME*PAID + GENDER*TIME*PAID + GROUP*GENDER*TIME*PAID

      2. Be careful to quote unusual variable names, (i.e., variables with embedded blanks, "stars" (*), etc).

  2. If the input Spreadsheet does contain a previously-saved default design, then

    1. You do not need to select variables for the analyses or specify the design and design effects. The design information stored with the spreadsheet will be used to conduct the analyses, although you can modify the design and the random effects.

    2. The Variables button on the Quick tab of the ByGroup: Variance Estimation and Precision dialog will be dimmed. The selection of dependent and categorical (grouping) variables can only be changed using the interactive Variance Estimation and Precision module.

    3. You can change the design and/or random effects, using the Edit options on the Advanced tab of the ByGroup: Variance Estimation and Precision dialog. When making changes, be sure to use the guidelines given in 1.C, above.

For more information on ByGroup Analysis, see ByGroup Analysis; see also, the ByGroup Analysis Example and the ByGroup Specifications for Variance Estimation and Precision.