This data representation does not require any assumption about the underlying statistical distribution and is based in the quartile division of the data.
A quartile division of linear and non linear scores was performed in order to identify 4 classes of predicted EWL for both models, on the basis of the values of age and selected psychological variables (Fig. 4 5).
The quartile division was performed on the entire set of data with no information regarding the grouping of samples into cancer cases or controls.
Quartile division was not possible for the index of chronic damage because a large number of biopsy specimens showed an index of 0%.
For some analyses, lifetime heroin exposure was cut into quartile categories with divisions at 0, 0.01, 2.49, 6.59 and 350 g-years, respectively, yielding 583, 207, 226 and 227 patients in each group, respectively.
These divisions were determined on age quartiles in the data, which provided appropriate groupings to reflect structural and functional changes in children related to their growth and maturation [ 21].
Body mass index and level of physical activity were included in models after division into quartiles.
In univariate and multivariable logistic regression analyses the continuous covariates were analyzed as categorical variables, after division into quartiles, to avoid assumptions of linearity in the log odds and to minimize the effect of outlying values.
Our scoring was primarily adapted from IWIH, i.e. division into quartiles, but for tumors with less than 25% stained cells a distinction was made between those with 6 25% positive tumor cells, those with approximately 1 5% positive tumor cells and those with absolute no HLA class I positive-stained tumor cells.
While in the UpQu normalization method, counts are normalized by division (in a given replicate) by the upper quartile of these counts, the Medi simply computes the median, and the ToCo uses the sum of all counts.
