Robust non-parametric one-sample tests for the analysis of recurrent events

Area of Science:

• Biostatistics
• Epidemiology
• Medical Statistics

Background:

• Recurrent events are common in medical studies, influencing treatment efficacy and safety assessments.
• Existing statistical methods may not adequately address the complexities of recurrent event data, especially in rare diseases.
• Accurate inference on recurrent events is crucial for evaluating interventions and understanding disease progression.

Purpose of the Study:

• To propose novel one-sample non-parametric tests for analyzing recurrent events.
• To develop robust and stratified versions of these tests to accommodate various hypotheses.
• To evaluate the performance of these tests under diverse event generation processes.

Main Methods:

• Development of one-sample non-parametric tests based on the standardized distance between observed and expected event counts.
• Inclusion of different weighting schemes to address various alternative hypotheses.
• Simulation studies to assess test performance under homogeneous/nonhomogeneous Poisson, autoregressive, and renewal processes, with and without frailty.
• Investigation of robust and stratified test versions.

Main Results:

• The proposed non-parametric tests effectively handle recurrent event data.
• Robust versions demonstrated suitability across a wide range of event generation processes.
• The tests provide a basis for inference on the marginal mean function of recurrent events.
• Simulation results confirmed the performance of the developed tests.

Conclusions:

• Robust non-parametric one-sample tests are valuable tools for analyzing recurrent events.
• These tests can be applied in non-randomized or epidemiological studies to assess treatment efficacy and safety, particularly in rare conditions.
• The methods offer a flexible approach for inference on recurrent event processes.