Modifying designs



This help file describes the menu for modifying designs


Freshly-generated designs do not have any responses. For very small designs, response values can be added by using the data editor within the R-commander. For larger designs, it is preferrable to do all editing outside of R and re-combine response data with the structural information using the import menu. It is also possible to add a response from a vector or data frame within R by using the menu item Add response variable(s).

Default analyses for experimental designs either work with all responses or with the first response of the design. It can therefore be advantageous to delete responses from the list of responses. This can be done with the menu item Select/Deselect response variables. At a later stage, one might want to redefine previous response variables as responses. This is also possible via that menu.

Contrary to response deselection, which leaves the variable in the design, one may want to permanently remove a column (response or other variable, but not experimental factor or block factor), for example because some artificial data for playing are no longer needed with real response data available. This can be done with the Remove column(s) menu item.

Qualitative design factors come with prespecified contrast coding, which influences how analysis results appear. If the default contrasts are not the desired ones, they can be changed from the Change contrasts menu item. Note that this is preferrable to the built-in function from R-Commander, because it preserves the integrity of the class design object.

Some designs come in long format (repeated measurements, some parameter designs) but should be brought into wide format and aggregated for conducting simple analyses. The reformatting can be done with the menu item Change from long to wide format, the subsequent aggregation with the menu item Aggregate design.

The menu item Fold design (not yet implemented) will allow to augment a 2-level factorial designs with a second portion as large as the first one by reversing the levels of some or all factors.

The menu item Add center points allows to add center points to 2-level designs with only quantitative factors and certain further restrictions (e.g. no splitplot design).

The menu item Add star portion for central composite design allows to add a starpoints with center points to regular (fractional) factorial 2-level designs with only quantitative factors and certain further restrictions (e.g. no splitplot design, no estimability requirements).

The menu item Augment lhs design allows to add additional points to latin hypercube sample designs according to different types of augmentation rules available in R-package lhs.

Note that command line use of the packages underlying this R-Commander routine is more flexible and allows more tuning.


Ulrike Groemping

Want to suggest features or report bugs for Use the GitHub issue tracker.