New approach and new program for analyses of false negatives-contaminated data in medicine and biology

BackgroundNo serological assay has 100% sensitivity. Statistically, the concentration of specific antibodies against antigens of parasites decreases with the duration of infection. This can result in false negative outputs of diagnostic tests for the subjects with old infectiong, e.g., for individuals infected in childhood. When a property of seronegative and seropositive subjects is compared under these circumstances, the statistical tests can detect no significant difference between these two groups of subjects, despite the fact that infected and noninfected subjects differ. When the effect of the infection has a cumulative character and subjects with an older infection (potential false negatives) are affected to a greater degree, we can even get paradoxical result of the comparison - the seronegative subjects have on average lower value of certain traits, e.g. IQ, despite the infection having a negative effect on the trait. A permutation test for the contaminated data, implemented, e.g., in the program Treept or available as a comprehensibly commented R function in the supplement of this paper, can be used to reveal and to eliminate the effect of false negatives.\n\nMethodsWe used a Monte Carlo simulation in the program R to show that the permutation test implemented in the programs Treept and PTPT is a conservative test.\n\nResultsWe showed that the test could provide false negative but not false positive results if the studied population contains no subpopulation of false negative subjects. We also introduced R version of the test expanded by skewness analysis, which helps to estimate the proportion of false negative subjects based on the assumption of equal data skewness in groups of healthy and infected individuals.\n\nConclusionsBased on the results of simulations and our experience with empirical studies we recommend the usage of permutation test for contaminated data whenever seronegative and seropositive individuals are compared.