Experimental Design

Abstract

Statistics can be manipulated by other professions; it is critical that scientists follow strict rules in the use and application of statistical techniques. The use of statistics becomes our referee to help us decide if our ideas and suppositions are correct. If we design our experiments intelligently, collect our data accurately and analyze them correctly, we can determine if our experimental treatments produce clear effects. Statistical analysis does not provide 100% certainty, but it does provide us an objective basis to draw conclusions based on recognized techniques and accepted guidelines rather than mere hype or speculation. A statistically significant difference does not necessarily mean that the treatment will have a practical effect. For example, a small, but statistically significant color change may or may not affect consumer acceptability of a chocolate pudding as the typical consumer may not be as sensitive to color differences as a colorimeter or trained sensory panel. Likewise, if no statistical significance is found in the development of an off-flavor, we conclude that there is no significant difference in the experimental treatment. There may be consumers that can detect the specific off-flavor, but the general population shows no effect.

Answers to Questions Raised in Table 6.1

It is obvious that there are three factors as represented in the first three columns with a total of 12 treatments (2  ×  2  ×  3) in a balanced design. The next seven columns are responses to the treatments. Seven responses suggest that they represent days of the week. If that is so, day of the week would also be another factor, making it 84 treatments (2  ×  2  ×  3  ×  7). With 84 treatments, there is no replication. All responses are rounded off to the nearest 0.5 and the range is 4.5 to 8.0. To be analyzed statistically, the entire experiment would need to be replicated.