The Importance of Being Random

Home > The Importance of Being Random > Page 2

An observational study is a study where the investigators observe subjects and measure variables of interest without assigning treatments to the subjects. The treatment that each subject chooses/receives is determined beyond the control of the investigator. Marissa is conducting an observational study. She is not assigning specific items for the students to choose from, and she is merely observing what the students eat.

There is, however, a flaw or bias in Marissa’s study. Bias is when the design of a statistical study results in favored outcomes. For example, in Marissa’s study, she only observes students who bring their lunch to school. She does not know who is really packing the lunches. Thus, her observation could be biased towards students bringing more healthy snacks because the parents could be the ones actually packing the lunches.

A sample survey is a study where investigators survey subjects and measure variables of interest based on what the subjects report. Andrew is conducting a sample survey. He is asking the students which they think is better, having an apple or having a candy bar.

Andrew’s design is also biased. In Andrew’s survey, he is asking students which they think is better. Better is a vague word that could be interpreted differently by each subject who participates in the survey. For example, one student may think better means “better for you” while another may think better means “better tasting.” Also, the survey question does not directly test the question our friends are looking for, which was “does the average teenager prefer to eat healthy snacks or unhealthy snacks.”

In an experimental study, investigators apply treatments to experimental subjects and then observe the effect of the treatments on the experimental units. Allyson allows her test subjects to choose which they prefer, an apple or a candy bar. This is an experiment because she is applying the food to the subjects and observing which they select.

Allyson’s study is also biased. A teenager may prefer healthy snacks, but not like apples. A teenager may prefer to eat unhealthy snacks, but be allergic to chocolate. The test subjects may have already purchased (or brought) a healthy item for their lunch and desire a dessert to go with it, thus selecting a candy bar from her. In either case, the results of the study are inconclusive.

Lastly, it is important to note that all three of the friends’ studies lack randomization. They are choosing who they want to have participate in the study rather than randomly selecting the participants. This produces bias. For example, Allyson is giving away free food so she may only select people she likes. Andrew is handing out surveys and may subconsciously hand the survey out to athletes (presuming they eat healthy foods). Marissa only observes people who brought their lunch, excluding students who purchase their lunch. Randomly choosing participants for a study helps to eliminate potential bias in the results.

< Previous Next >