Evaluation of Forensic Data Using Logistic Regression-Based Classification Methods and an R Shiny Implementation

A methodological comparison of discriminant function analysis and binary logistic regression for estimating sex in forensic research and case-work

“…Similarly, specificity and negative predicted value are the true negative rates, but the negative predicted value is based on the sum total of the number of true negatives and false negatives, whereas the specificity is based on the sum total of the number of true negative and false positives, where the true negative is the frequency of correct prediction of males when they were actually males, true positive is the frequency of correct prediction of females when they were actually females, false negative is the frequency of false prediction of females when they were actually males, and false positive is the frequency of false prediction of males when they were actually females. 17,33…”

“…The value of CPP ranges from 0% to 100%, with 100% being the best prediction, whereas the value of classification error varies from 0 to 1, with 0 indicating the best prediction, that is, the lower the error, the more reliable the model. 17 Additionally, classification error is regarded as an inappropriate comparison metric, because it is usually overestimated, that is, it does not effectively reflect the differences in values. 22 Moreover, the Matthews correlation coefficient incorporates all the four elements, namely true positive, true negative, false positive, and false negative, and hence has an advantage over other accuracy measures such as positive predictive value, negative predictive value, and so on.…”

“…The sensitivity (also known as true positive rate) of a binary classification test concerning a class is the probability that the test correctly classifies cases belonging to that class, whereas the specificity (also known as true negative rate) is the probability that the test correctly classifies cases not belonging to that class 17 :The positive predictive value (PPV) also called precision is defined as how correctly the observed characteristics were predicted as compared to the total number of cases predicted as having characteristics. Whereas the negative predictive value (NPV) is defined as the correct prediction of cases without the observed characteristics as compared to the total number of cases predicted as not having the characteristics.…”

“…In contrast, the B index calculates the average of the squared difference between the estimated and actual values. The formulae for the computation of CPP and B index were followed by Liong and Foo 21 and are listed below.where P i = probability of classification into group i Y i = actual group membership (male or female) n = sample size of both the populations Moreover, the classification error is the rate of observations in a group that are erroneously classified, that is, the total number of false positives and false negatives related to the total number of cases 17 :Additionally, the Matthews correlation coefficient (MMC) was calculated. The below-mentioned formula was used to calculate MMC, which was followed by Vihinen 33 :where TP = True Positives, TN = True Negatives, FP = False Positives, FN = False Negatives.…”

“…Along with it, deciphering the sophisticated statistical analysis and presenting them in a way that the judicial system can easily comprehend is another essential responsibility of a forensic expert. 17…”