Botched Designs

So-called "botched" designs are experiments that include experimental runs where the factor settings are not set precisely to the respective level(s) as intended and as consistent with the original design. STATISTICA Experimental Design can correctly analyze botched designs for 2-level factors, with or without center points (botched designs for experiments with factors at more than 2 levels are not supported in the Experimental Design module but can still be analyzed using the General Linear Models features of STATISTICA).

In general, experimental designs for 2-level factors are usually constructed so as to extract as much information as possible from as few runs (cases) as possible; this is described in the Introductory Overviews for Experimental Design, in particular the sections on 2(k-p) Fractional Factorial Designs and 2(k-p) Maximally Unconfounded and Minimum Aberration Designs, as well as 2-level Screening Designs. It is not uncommon, however, that when the actual data are collected for the experiment, some of the exact settings (called for by the specific experimental design) cannot be achieved and as a result, they are only approximated (e.g., instead of a value 100, a value 99 is substituted for the respective variable and run). Such imprecise settings will alter the properties (confounding, resolution) of the experiment; however, unless the experimental runs are grossly mismatched (different from the desired values), valid and meaningful analyses can still be conducted.