Chaughule, Thorat - Statistical Analysis/Design of Experiments
8.7. FREQUENT DESIGN TROUBLES
In the experimental design, statistical analyses of the results are often inconclusive
or, misleading if statistical considerations are not incorporated. Below are few of many
potential problems that can occur when statistical methodology is not used to design
drying or food engineering experiments.
8.7.1. Experimental variation masking factor effects
Researchers often invest substantial project funds and a great amount of time and
effort only to find that the research hypotheses are not supported by the experimental
results. Many times the lack of statistical confirmation is the result of the inherent varia-
bility of the test results.
For example, color kinetics during drying of Sapota fruit there can either be a varia-
tion among different test runs due to experimental procedures or variation in the raw
material source and even maturity of the fruit can change the color estimation. Experi-
ments that are intended to study factor effects (e.g., drying time) on fruit color must be
designed so that the variation in subjects and across time does not mask the effects of
the experimental factors.
It should be noted that, the experimental error variation must be considered in the
statistical design of an experiment. Failure to do so could result in true factor effects be-
ing hidden by the variation in the observed responses.
8.7.2. Uncontrolled factors compromise experimental conclusions
Sometimes researchers would intentionally ignore factors that are known to exert
important influences on a response; there are many subtle ways in which failure to care-
fully consider all factors of importance can compromise the conclusions drawn from ex-
perimental results.
Consider, for example, an experiment conducted to compare color changes in L, a, b
values during hot air drying of Sapota pulp. Suppose that an adequate experimental de-
sign has been selected, including a sufficiently large number of test runs. Figure 8.4 ex-
hibits actual results of such a test run. One of the important conclusions that can be
drawn from an analysis of the experimental data is that the L, a, b values decrease
slightly with respect to drying kinetics and then slight increase at the end of drying.