PersonAlytics: PersonAlytics: Analytics for single-case, small N, and...

In ICTatRTI/PersonAlytics: Analytics for N-of-1 and small N intensive longitudinal designs, idiographic clinical trials (ICT), and interrupted time series

PersonAlytics: Analytics for single-case, small N, and Idiographic Clinical Trials:

Description

The PersonAlytics package provides the simplified user interface for implementing linear mixed effects models for idiographic clinical trials (ICT) data, single case studies, and small N studies with intensive longitudinal designs. Contact us via https://personalytics.rti.org/ for licensing options.

Details

The basic mixed effects model is dv=time+phase+phase*time with random intercepts and random slopes for time. The phase variable is optional. Additional independent variables (or covariates) can be included. The PersonAlytics package provides the simplified user interface for implementing this model using gamlss or lme. The primary function of PersonAlytics is PersonAlytic.

Key features of the PersonAlytics package include:

Automated detection of the residual covariance structure. PersonAlytics automates model comparisons for determining autocorrelation structure for all patients or for each patient.

Automated detection of the function form for the time variable. PersonAlytics automates model comparisons for determining the functional form of the relationship between time and the outcome (i.e., linear vs. quadratic vs. cubic growth models) for all patients or for each patient.

Estimation. The automated covariance structure and function form for time is done using maximum likelihood (ML) estimators. Final results are estimated using restricted maximum likelihood (REML).

High Throughput. When users have a list of outcomes (dependent variables), a list of target covariates, and/or or desire the analyses to be repeated for each individual in the data set, high throughput options automate the model fitting process.

False Discovery Rate Adjustment. When high throughput options are requested, Type I error correction and false discovery rate adjustments are implemented post-implementation across target covariates (and individuals if requested) within each outcome.

Linear and Generalized Linear Mixed Effects Models. Linear mixed effects models can be fit in either the nlme framework or the gamlss approach. The two approaches give nearly identical fixed effects estimates but differ in their computation of standard errors and random effects. Generalized linear mixed effects models can be fit using the gamlss option (see gamlss.family). The gamlss approach also allows models for dealing with heteroscedasticity implemented by including mixed effects models for the variance.

The Palytic class

See Palytic

Author(s)

Stephen Tueller [e-mail]

ICTatRTI/PersonAlytics documentation built on Dec. 13, 2024, 11:06 p.m.