Trayless week: an inconclusive inconvenience

The front-page article printed in Chips (Vol. 130, No. 21) regarding trayless week was misleading, to say the least. Dealing with statistics often is, as it can be difficult to know what exactly your data is telling you. There are a number of issues that must be addressed to draw any sort of reasonable conclusions that this study failed to look at in any way.

The first obvious flaw in the study was that there was no control over the meals served. Different meals cause different amounts of waste. As Chips noted, one of the meals measured had chicken bones and watermelon rinds included in its measurements, and many other things like particularly messy meals resulting in more napkins thrown away can impact the amount of waste produced at a meal. Another factor that the study actually did try to account for — but was more or less ignored in the article — was that the number of people eating has an impact on the amount of waste produced. This data was included in the table summarizing the data, but the 8.4 percent reduction the article referred to was derived from the total amount of waste produced in the two weeks, not the amount of waste per person.

Perhaps one of the most important things ignored in this study was that it only spanned 10 meals. It would be absurd to believe that people who stub their toes before a history test perform better because a study was done in which five people who stubbed their toes did slightly better than five who did not. This is known as sampling error, which statistical studies usually mitigate by being big enough to represent the population they relate to. The smaller the study, the more likely it is that random flukes will lead to misleading conclusions.

With studies of any size, statisticians create regression models to analyze the data. These models include a response variable and a number of predictor variables. In this study the predictor variable is the presence or absence of trays, which attempts to predict the response variable of waste per person. Creating and analyzing this model shows that there is a 60 percent chance that removing trays does not change the amount of waste produced at all. Generally a variable is not considered significant until there is less than a five percent chance that it doesn’t have an impact on the response variable. We can also say that it is 95 percent likely that removing trays does anything from reducing the amount of waste per person by .09 pounds to actually increasing the amount of waste produced by .06 pounds per person. In other words, the study, flawed to begin with, does not statistically support the hypothesis that removing trays from the cafeteria reduces waste.

However, before people rush off to do the exact same study over again over a longer period of time, they should really try to study the causes of waste in the cafeteria. Removing trays was a study that looked at a solution to waste in the cafeteria, but it did not seem to have much foundation. It would make a heck of a lot more sense to look at the causes of waste in the cafeteria first: what reasons people have for wasting food and what meals, meal times and days produce more waste, among other things. Based on that information a reasonable solution could be developed and tested. And for goodness’ sake, talk to a statistician first.

Andrew Tharp (‘08)