In natural science and signal processing, an artifact is any perceived distortion or other data error caused by the instrument of observation.
For example, a compression artifact in computer science is a noticeable error caused by lossy data compression. In microscopy, artifacts are sometimes introduced during the processing of samples into slide form.
In econometrics, which trades on computing linear relationships between related variables, an artifact is a spurious finding, such as one based on either a faulty choice of variables or an overextension of the computed relationship. Such an artifact may be called a statistical artifact. For instance, a hypothetical finding that presidential approval rating is approximately equal to twice the percentage of citizens making more than $50,000 annually would predict that the approval rating will be 120% if 60% of citizens make over $50,000. This prediction is a statistical artifact, since it is spurious to use the model when the percentage of citizens making over $50,000 is so high (and silly to predict an approval rating greater than 100%).
