|
Abstract:
|
The genetic code can be considered as nominal data, since there is no known way to quantify each character. Therefore, when we evaluate an association between two positions, we usually rely on test statistics derived from a contingency table that asymptotically follow a chi^2-distribution. Such test statistics are by nature two-sided. To develop a one-sided method, we note that genes tend to be very conservative, and as a result, viable mutations are so rare that usually only one dominant character (the consensus) is allowed at each position. Exploiting this fact, we develop a one-sided test to evaluate the association with respect to the direction away from the consensus. Appropriate test statistics are presented with an exact method for assessment of statistical significance.
|
