Partitioning the Sums of Squares (2 of 7)

next previous
Error
All other differences among the subjects' scores are said to be due to "error." Do not take the word "error" literally, however. It does not mean that a mistake was made. For historical reasons, any differences among subjects that cannot be explained by the experimental treatments are called error. It would be better, perhaps, to call this source of differences "unexplained variation," but to be consistent with the terminology in common usage, these differences will be called "error."

A major source of error is the simple fact that even if subjects are treated exactly the same way in an experiment, they differ from each other, often greatly. This stands to reason since subjects differ in both their pre-experimental experiences and in their genetics. In terms of this example, it means that subjects given the same drug will not necessarily experience the same degree of pain relief. Indeed, there are differences among subjects within each of the three groups.
next previous