Analyzing and Presenting pH Data

Both the practical effects and theoritical basis of error in data analysis and presentation were examined. Since the pH transformation has been applied for convenience and not for statistical reason, added care must be exercised in analyzing these data. Calculating a mean of observed pH's introduced a significant and predictable bias as compared to −log (mean [H + ]), henceforth designated mH. Rumen data from Holstein steers fed a high concentrate diet twice daily showed that a relative error of from .35 to 2.22%, with an average of 1.25%, resulted when the former method was used. Wide ranges in observed or low greatly increased the possible relative error. Statistical analyses of data should use the untransformed variable, i.e., hydrogen ion concentration. Mean hydrogen ion concentration, the variable in which researchers are interested, still can be transformed for convenience and reported as mH together with the asymmetric confidence intervals that result from transformation.