Useful Functions for Ridit Analysis (ridittools)
Functions to compute ridit scores of vectors, compute mean ridits and their standard errors for vectors compared to a reference vector, as described in Fleiss (1981, ISBN:0-471-06428-9), and compute means/SEs for multiple groups in matrices. Data can be either counts or proportions. Emphasis is on ridit analysis of ordered categorical data such as Likert items and pain-rating scales.

Cross-Recurrence Quantification Analysis for Dynamic Natural Language Processing (crqanlp)
Cross-recurrence quantification analysis for word series, from text, known as categorical recurrence analysis. Uses the ‘crqa’ R package by Coco and Dale (2014) <doi:10.3389/fpsyg.2014.00510>. Functions are wrappers to facilitate exploration of the sequential properties of text.

Spatial Feature Smoothing (smoothr)
Smooth spatial features (i.e. lines and polygons) to remove sharp corners and make curves appear more natural or aesthetically pleasing. Two smoothing methods are available: Chaikin’s corner cutting algorithm (Chaikin 1974 <doi:10.1016/0146-664X(74)90028-8>) and spline interpolation.

Measure a Subject’s Abnormality with Respect to a Reference Population (abnormality)
Contains the functions to implement the methodology and considerations laid out by Marks et al. in the manuscript Measuring Abnormality in High Dimensional Spaces: Applications in Biomechanical Gait Analysis. As of 2/27/2018 this paper has been submitted and is under scientific review. Using high-dimensional datasets to measure a subject’s overall level of abnormality as compared to a reference population is often needed in outcomes research. Utilizing applications in instrumented gait analysis, that article demonstrates how using data that is inherently non-independent to measure overall abnormality may bias results. A methodology is introduced to address this bias to accurately measure overall abnormality in high dimensional spaces. While this methodology is in line with previous literature, it differs in two major ways. Advantageously, it can be applied to datasets in which the number of observations is less than the number of features/variables, and it can be abstracted to practically any number of domains or dimensions. After applying the proposed methodology to the original data, the researcher is left with a set of uncorrelated variables (i.e. principal components) with which overall abnormality can be measured without bias. Different considerations are discussed in that article in deciding the appropriate number of principal components to keep and the aggregate distance measure to utilize.

Time Series Forecasting Using Nearest Neighbors (tsfknn)
Allows to forecast time series using nearest neighbors regression Francisco Martinez, Maria P. Frias, Maria D. Perez-Godoy and Antonio J. Rivera (2017) <doi:10.1007/s10462-017-9593-z>. When the forecasting horizon multi-step ahead forecasting strategies can be used. The model built is is higher than 1, two autoregressive, that is, it is only based on the observations of the time series. The nearest neighbors used in a prediction can be consulted and plotted.

Advertisements