Charting temperature effects on lab readings

Hy Sedrate, a quality assurance specialist for St. Recover in the Long Run Hospital, has noticed that the accuracy of his lab readings is higher in cool weather than on hot days, even though the lab is air conditioned and does not seem to have an appreciable difference in temperature from that of the rest of the hospital. Nonetheless, the data is dramatic, and Hy wants to demonstrate the pattern that he has observed so that his boss and the entire quality team can analyze the situation.

Hy collects data relating to the number of defective lab readings, and at the same time keeps track of daily high temperatures recorded by the National Weather Service for his area. Without complex analysis, he can see even from the raw data that the higher the temperature, the greater the number of errors. Nonetheless, in order to illustrate the pattern that he has observed, Hy decides to create an individual moving range chart from the data.

After he has produced the charts and examined them for patterns, another quality specialist, Hap N. Stance, insists that Hy has used the wrong statistical method, although Hap cannot advise what approach might be better. Is Hy Sedrate correct in using an individual moving range chart, or is there a better way to analyze this data?

A) Hy should abandon his chart and take an Alka Seltzer. It's too hot outside to think.

B) An individual moving range chart provides insights about the data that no other method can give.

C) Hap N. Stance is on the right track, but the wrong train. He should advise Hy to use a scatter diagram to get the best analysis of the data.

D) Since he already knows that there is a relationship between defects and temperature, Hy should select the chart that will be most dramatic, so he can impress the quality team with his statistical prowess.

C is the correct answer.

A scatter diagram will indicate whether two variables-in this case, temperature and inaccurate lab reports-are related to each other, or whether there is a correlation between the two. If one of the variables appears to have an effect on the other, then regression analysis will be appropriate. Since Hy Sedrate wants to establish whether there is indeed a correlation, an individual moving range chart will not be of any help here. (See Practical Tools for Continuous Improvement (Volume I, Statistical Tools), Jacqueline D. Graham, Ph.D., and Michael J. Cleary, Ph.D., pp. 298-311).

The same data can be charted using CHARTrunner from PQ Systems as follows:

Since Hy Sedrate assumes that temperature affects the number of defects (rather than the other way around), temperature becomes the independent variable, represented on the horizontal axis. The number of defects becomes the dependent variable, on the vertical axis. The line of best fit shows the relationship:

Y = 13.44 + 1.66X

This means that as the temperature goes up, so does the number of defects. A correlation coefficient of .8 suggests a rather strong relationship between the two variables.