Nuisance factors

Variations observed in gene expressions may be attributable to differences in the conditions being compared in the samples (for example, normal and disease tissues), or to differences in nuisance factors such as array, dye, and spot. Thus, we believe the design of microarrays should follow the statistical principles of blocking, randomization, and replication. 

Randomization is the basis for valid statistical inference. Random assignment of replicate samples to arrays and dyes helps to avoid unintended systematic bias due to such nuisance factors. In designing a two-channel microarray, a software driver for the robotic arrayer allows appropriate programming to randomize and replicate the spotting of the complementary DNA (cDNA) on each array. Unfortunately, randomization of spotting on each array is often not done currently, perhaps due to inconvenience. 

The presence of solids, even in smell quantities, can be an extreme nuisance factor in a system design. Solids can readily foul packed columog. mesh pads, or fiber beds unless prescrubbing or significant irrigation is used to flush contact surfaces. 

The benefit of replication is that the average of several observations comes closer to the true vedue than a single observation. Replication helps bedemce out the bias due to the effect of nuisance factors. It also helps to detect gross errors in the measurements. It therefore improves the precision of the statistical inferences. 

It is extremely important to note that false negatives, for whatever reason, from a detector could prove to be fatal if the device cannot detect targeted toxic substances. Therefore, research emphasis should not only focus on reducing the false positive alarm rate. The potential for false negatives is more important and must be addressed. While false positives are nuisance factors, false negatives could lead to fatalities. 

In many experimental design problems, it is necessary to design the trials or runs in such a way that the variability arising from some nuisance factors can be controlled. These nuisance factors (or "blocking variables ) may affect the response but are neither factors nor fixed variables. They can be analysts, instruments, reagents, sample materials, working day, etc. Let us consider the simple case of a 2 factorial design to improve an automated extraction step in a... 

---