dc.description.abstract | Robust likelihood inferences for comparing the semi-positive values of diagnostic procedures in paired designs
Abstract
In cancer screening, not only the disease is concerned, but also the location of the tumor is very important. Therefore, Lim et al. (2018) proposed semi-positive predictive value and semi-sensitivity two new indicators, and based on joint distribution in paired designs, using the statistical method of Bennett (1972), a chi-square test statistic comparing the semi-positive predictive values and semi-sensitivity of the two screening tests was proposed.
This paper presents a method for comparing the semi-positive predictive values of two screens with a robust likelihood function. Compared with the above-mentioned joint distribution, we assume data independence, that is, ignore the correlation, and use the working model under independence. For the parameter of interest, that is, the difference between the semi-positive predictive values of the two screens, we robustify the actual likelihood function and propose a robust statistic based on the correct likelihood function. We utilize simulations to compare our statistic and other current existing test statistics. | en_US |