1 April 1998 Spurious Replication in Analysis: Making 2 × 2 Tables out of Paired Data
Arthur W. Ghent
Author Affiliations +

An example is cited from a published study of forceps (cerci) asymmetry and copulatory success in earwigs (Forficula auricularia). Its authors scored eight pairs of male earwigs—one with symmetric and one with asymmetric forceps—according to which male first succeeded in mating with a female earwig enclosed with them in a petri dish. Their categories “symmetric males successful” (seven observed among the eight pairs) and “asymmetric males unsuccessful” (the seven pairmates of the successful males) were thereby locked together like the observations “seven Heads up” and “seven Tails down” in tosses of eight coins. The authors assumed, however, that their eight petri dishes were providing two independent observations per dish. They analyzed this assumed total of 16 observations by the Fisher exact test of a 2 × 2 table, obtaining an apparent two-sided probability of 0.01. In fact, the study provided only eight independent observations, correctly testable by a binomial exact test of the 7:1 ratio. By this appropriate analysis, observed counts of 7 to 1 fail to achieve conventional significance (0.05) at a two-sided probability of 0.07.

Arthur W. Ghent "Spurious Replication in Analysis: Making 2 × 2 Tables out of Paired Data," The American Midland Naturalist 139(2), 391-393, (1 April 1998). https://doi.org/10.1674/0003-0031(1998)139[0391:SRIAMT]2.0.CO;2
Received: 12 January 1997; Accepted: 1 September 1997; Published: 1 April 1998
Get copyright permission
Back to Top