A Practical Method for Remedying Some Biased Data Sets: Looking for Fractal Traces

Abstract

Background and aim. Biases are largely diffused in medical studies. Methods for remedying biases in medical literature would be advisable. Here we propose a simple tool for correcting some biased data by calculating the proportion of non-random biases embedded in the variance. Method. Starting from a variable in a given series, the proportion of variability due to fractal behavior and the proportion of variability due to stochastic distribution can be calculated. Thus, the ratio among proportions can provide the amount of variance due to non-random biases. This proportion is used for re-calculating standard errors and confidence intervals. The method is applied to 2018 rates of in-hospital births in the Umbria region of Italy and to a set of effect sizes re-calculated from already published systematic reviews and meta-analyses on intrahepatic cholestasis of pregnancy. Those topics have been chosen because they have been already acknowledged as heterogeneous. Results. Proofs demonstrated that corrected standard errors work better than usual standard errors in heterogeneous data syntheses. Conclusion. By combining fractal behavior and stochastic distribution characteristics to data sets, some non-random biases can be corrected.