We are constantly inundated by a sea of information, and in the age of the internet,  every claim needs to be backed up by several sources to even have a chance at being considered reputable. But statistics and research aren’t always what they seem.

It’s incredibly simple to manipulate statistics — percentages and charts feel intuitive.  Unverifiable claims like “nine out of 10 dentists recommend … ” are ubiquitous. However, there are several more sophisticated and insidious techniques to manipulate numbers that are much harder to identify.  

One easy way to misuse data is by applying the conclusions of a study to a population it did not accurately sample. For example, many medical studies and clinical trials center on white people, while underrepresenting or even excluding minority groups altogether. While the results of the study may very well be useful in treating and diagnosing the population of white males, it is not accurate and sometimes even actively harmful to claim and make decisions as if the study gives a holistic portrayal of how a disorder presents in all people.  

While errors like this are sometimes a result of genuine ignorance, the consequences of making claims about a population based on a limited or biased sample are overwhelmingly negative, and culpability for those consequences lies on the shoulders of researchers. 

Apart from errors of convenience or bias, there are many corporate and government-funded studies  designed to seek a specific conclusion; in other words, the study is likely to produce results in the interests of the organization funding it. This was a tactic infamously used by Big Tobacco, a group that not only funded and published research that aligned with their interests but actively suppressed and criticized opposing research.  

By carefully choosing a sample, and by tweaking the constraints of a statistical model that delineate the line between a successs and a failure, inconclusive data can be nudged in the direction of a favorable outcome. Furthermore, the choices made when visualizing data can also lead to deception. From the size of the ticks to the colors used to using a backward scale, it’s almost too easy to make boring data hit hard and make shocking results seem commonplace. 

Every researcher needs to be aware of these pitfalls. Those seriously pursuing discovery and truth must be acutely mindful of drawing conclusions truly supported by the data. It’s an impossible task to look into every statistic you come across, but simply understanding bias, error and room for interpretation are qualities inherent to statistics is a big step up from accepting statistics at face value.

Sandhya is a junior in LAS.

[email protected]