Errors Type I/Type II Errors

¶ … Errors

Type I/Type II Errors

Statistical analysis can lead to many different errors of many different types, both in the gathering of data and the manipulation of it to produce results in a practical and relevant manner. Often, errors arise as a result of the complex mathematical manipulations that must occur in order to make useful sense of data. These mathematical errors can compound and lead to wildly incorrect interpretations of data, producing results that cannot be trusted or validly used. Other errors can occur in the interpretive phase of data analysis; these can often be far more egregious, and at the same time they are often more difficult to catch. Errors made in that actual mathematic manipulation of data often delivers results that -- for obvious reasons -- simply do not make sense. Interpretive errors, however, are more difficult to catch almost by definition. The data itself may be entirely sound, and therefore the results are more likely to be trusted, but an error in interpretation can still cause the data to be incorrectly applied.

There are two rather basic and fairly straightforward errors, known as Type I and Type II errors, that are commonly made in data analysis. Both refer to a basic mistake regarding the status quo from which the analysis is meant to measure change. This status quo is called the null hypothesis, the idea/belief that there was no change in the phenomenon measured during the test. When there is no change in the situation or phenomenon, the null hypothesis is said to be true (that is, nothing happened). If a change in the situation/phenomenon has in fact occurred, then the null hypothesis (the idea that nothing has happened) is quite clearly false. A Type I error occurs when there is a false positive -- when the data analysis suggests a change has occurred, when in fact there has been no change. Thus, in a Type I error the null hypothesis is true but falsely rejected. A Type II error is the opposite -- the null hypothesis is false, but is accepted as true (a false negative).

One commonly used (and perhaps commonly experienced) Type I error occurs in the use of home pregnancy tests. When the test returns a positive result (meaning that the woman is indeed pregnant), but in fact no pregnancy exists, the test has returned a false positive, meaning a Type I error has occurred. In this instance the null hypothesis would be no pregnancy -- no change from the status quo. The null hypothesis that there is no pregnancy is falsely rejected when the pregnancy test returns a positive result without a pregnancy truly existing.