For the last Brain Teaser, in an interesting spin, sample size was defined as a time window. A is the correct response. The underlying data type falls into the category of defect counts (complaints) within a constant area of opportunity (one hour). These data, of course, follow the assumptions of a Poisson distribution, which are monitored on either c or u charts. Because the complaints are tallied every hour, one would use a c-chart. Group processing is needed because multiple process streams (three different products) are to be monitored on a single chart. Also, the data need to be normalized using short-run processing because the complaint rate is different among the three products. Short-run processing transforms the data to ratios of actual complaint count to expected count.

Don’t confuse complaint counts with sample size. The hourly complaint tally is the data value. The sample size makes the complaint count relative to an area of opportunity. In this case, our area of opportunity is one hour. We typically think of counting defects on a part, in a bag, from a pallet or from a lot, and in this example sample size is a time window.