The premise of this follow-up sister study to "Measuring the Frequency Occurrence of Handwriting and Handprinting Characteristics" was to collect a representative population sampling of numerals and assess how many participants utilize each of the predetermined characteristics as found in their specimens. A total of 1410 handwriting specimen forms were collected from across the United States and pared to 1025 to obtain a proper representative sample of the U.S. adult population based on the same demographics used in the original 2017 study. This study provides frequency of occurrence proportions and 95% confidence limits for 25 handwritten numeral characteristics. A total of 277 intercharacter pairs of handwritten numeral characteristics were cross-analyzed for interdependence. The results were that 72.92% of all intercharacter pairs had a coefficient of correlation between -0.2 and +0.2 in this study.

This report describes the results of a statistical research study. The purpose of this study, and its predecessors, has been, in part, to provide a statistical foundation to handwriting comparison science. With the establishment of the initial database of statistically valid frequency occurrence proportions for handwriting, hand printing and numerals completed, the research team now focuses on filling gaps within the database and expanding the database of writers. As these numbers increase, both the number of proportions and their statistical accuracy increase. The authors identified an initial set of 34 numeral characteristic features for this study. These features were selected based on their seemingly objective description and value for use by forensic document examiners in their casework. These features were tested for their level of objectivity through an attribute agreement analysis (AAA). Only those features of which there was unanimous agreement, 17 features, were admitted into the study database based on the objectivity indicated by the AAA testing. Handwriting specimens from 1410 participants had previously been collected and were utilized in this study. This number was pared to 1024 participant forms based on parameters originally

A global representation of polyp and non polyp shapes are constructed following a point distribution model (PDM) as an alternative to current methods which only inspect local shape characteristics at a point on the surface. The decision on whether or not a candidate lesion is a polyp can then be made by comparing the minimum Euclidean distance of the candidate lesion to the constructed mean shapes. The model closer in distance to the candidate lesion is selected to represent that particular lesion -polyp or non polyp. This shape model can also be used to investigate the shape variability of the different lesions detected by constructing an allowable shape domain for each of these lesions.

No abstract