Research Tips

Dealing with Outliers

Before undertaking rigorous data analysis, it’s a good idea to check your data for outliers. Outliers are values outside the normal or expected range, and they may occur for various reasons, including:

The data may have been coded or entered incorrectly.
The respondent may have misunderstood the question.
For some reason, the respondent is very different from other respondents in terms of whatever is being measured.

There are complicated methods of searching for outliers, but an easy starting point is to examine frequency distributions for unusual or unexpected values. Data coding and entry errors leading to outliers should be corrected, but dealing with other outliers requires careful consideration. In some situations, outliers can be eliminated to prevent them from unduly influencing results, but this should be done with caution if the outlying observations are valid but just different from most others.