No abstract

No abstract

No abstract

The merits of using the arcsine transformation prior to analyzing proportion data is being questioned in the published literature. While arcsine transformation stabilizes variance and normalizes proportional data, there are several reasons why this method can be problematic. An alternative analysis proposed to address the problems with normality of proportion data is the Generalized Linear Model logistic regression analysis. We compared the frequency of use of arcsine through time in ten leading biological journals. We tested the effectiveness of both arcsine transformation and logistic regression in making the residuals meet the assumptions of normality, homogeneity and independence by noting changes in the residual plots and changes in the p-value and significance decision compared to the linear regression on untransformed data using 40 data sets from the published literature. In the leading biological journals there is an obvious trend of an increased use of arcsine transformation on percentage data starting around the 1970s. Logistic regression was able to improve the residuals' normality, homogeneity and independence more often than arcsine. The arcsine transformation increased and decreased p values at almost the same rate. In comparison, logistic regression increased the p-value in 86% of the data sets, often resulting in a change in significance. The results suggest that logistic regression should be used as an alternative to the arcsine transformation in biological analysis.