What is a "p value" and how can it be used as a tool when identifying antibodies?

A "p value" is a statistical term found in scientific research papers in which two or more outcomes are compared. It can be defined as the probability that the results could have occurred by chance (sampling) alone. Applying p values to antibody identification is somewhat controversial (see #2 and #3 below).

Inferential statistics uses p values in which inferences are made about a population based on the sample studied and a p value is calculated to estimate the level of error that could exist due to chance sampling. Many statistical tests can be used to calculate p values.

As applied to antibody identification, a p value is calculated (using Fisher's exact method) to assess if an adequate number of red cells have been tested. In general, testing more cells decreases the p value and increases the likelihood that results did not occur by chance alone. A p value of 0.05 means that the same results caused by another antibody would be expected to occur by chance alone only one in 20 times (5% of the time). Setting the acceptable level of error at p=0.05 is arbitrary and flawed, but it reflects a widespread practice in science.

Caution: Calculating p values is just a statistical tool and does not substitute for carefully examining serologic test results. If acceptable, p values provide additional converging evidence to support the conclusion that the correct antibody has been identified.

To learn more about how to interpret p values, see Further Reading.