Statistical Methods for Life Data Analysis

Contains methods for examining bench test or field data using the well-known Weibull Analysis. It includes Monte Carlo simulation for estimating the life span of products that have not failed, taking account of registering and reporting delays as stated in (Verband der Automobilindustrie e.V. (VDA), 2016, ). If the products looked upon are vehicles, the covered mileage can be estimated as well. It also provides non-parametric estimators like Median Ranks, Kaplan-Meier (Abernethy, 2006, ), Johnson (Johnson, 1964, ), and Nelson-Aalen for failure probability estimation within samples that contain failures as well as censored data. Methods for estimating the parameters of lifetime distributions, like Maximum Likelihood and Median-Rank Regression, (Genschel and Meeker, 2010, ) as well as the computation of confidence intervals of quantiles and probabilities using the delta method related to Fisher's confidence intervals (Meeker and Escobar, 1998, ) and the beta-binomial confidence bounds are also included. If desired, the data can automatically be divided into subgroups using segmented regression. And if the number of subgroups in a Weibull Mixture Model is known, data can be analyzed using the EM-Algorithm. Besides the calculation, methods for interactive visualization of the edited data using *plotly* are provided as well. These visualizations include the layout of a probability plot for a specified distribution, the graphical technique of probability plotting and the possibility of adding regression lines and confidence bounds to existing plots.


weibulltools

Unlike other R packages for survival analysis, the weibulltools package focuses on reliability methods and has the advantage of being open source and easily accessible contrary to other reliability analysis software. In addition to that the package can be integrated into (partly) automated data analysis processes and even can be connected to big data systems.

The weibulltools package contains methods for examining bench test or field data using the well-known weibull analysis. It includes Monte Carlo simulation for estimating the life span of products that have not failed yet, taking account of registering and reporting delays. On this basis, if the products looked upon are vehicles, the covered mileage can be estimated as well. The weibulltools package also provides methods for probability estimation within samples that contain failures as well as censored data. Methods for estimating the parameters of lifetime distributions as well as the confidence intervals of quantiles and probabilities are also included. If desired, the data can automatically be divided into subgroups using segmented regression. And if the number of subgroups in a Weibull Mixture Model is known, data can be analyzed using the EM-Algorithm. Besides the calculation methods, methods for interactive visualization of the edited data using Plotly are provided as well.

News


title: "NEWS" output: html_document

Release of weibulltools v1.0.1

  • Fixed installation error when using clang compiler

Release of weibulltools v1.0.0

Prerequisite for Package Usage:

  • Since RcppArmadillo is used, the R version should be at least 3.3.0 (listed under Depends in DESCRIPTION file)

New Features:

  • Vignettes for non-parametric probability estimation, parameter estimation using Median-Rank Regression and Maximum-Likelihood and mixture model estimation are provided.
  • Argument y in functions plot_prob_mix() and plot_mod_mix() is deprecated and not used anymore.
  • Argument reg_output in functions plot_prob_mix() and plot_mod_mix() is deprecated; use mix_output instead.
  • Function plot_mod_mix() was revised and updated in the way that the obtained results of the function mixmod_em() can be visualized.
  • Function plot_prob_mix() was revised and updated in the way that the obtained results of the function mixmod_em() can be visualized.
  • Implementation of EM-Algorithm using Newton-Raphson. The algorithm is written in c++ (mixture_em_cpp()) and is called in mixmod_em().
  • New method for the computation of Fisher's Confidence Bounds regarding probabilities is used. These method is called "z-Procedure" and is more appropriate to manage the bend-back behaviour. Therefore an adjustment of functions delta_method() and confint_fisher() was made.
  • Implementation of log-location-scale models with threshold parameter like three-parametric weibull ("weibull3"), three-parametric lognormal ("lognormal3") and three-parametric loglogistic ("loglogistic3").
  • Implementation of location-scale models like smallest extreme value ("sev"), normal ("normal") and logistic ("logistic").
  • Implementation of Log-Likelihood Profiling for three-parametric models in function loglik_profiling(). In general this function is used inside ml_estimation() for the purpose of estimating threshold parameter of three-parametric models.
  • Implementation of R-Squared Profiling for three-parametric models in function r_squared_profiling(). In general this function is used inside rank_regression() for the purpose of estimating threshold parameter of three-parametric models.
  • Implementation of Log-Likelihood Function for all implemented models in function loglik_function(). In general this function is used inside ml_estimation() for the purpose of estimating the variance-covariance matrix of location-scale models "sev", "normal" and "logistic". The function is also used to estimate the variance-covariance matrix of log-location-scale models with a threshold parameter, i.e. "weibull3", "lognormal3" and "loglogistic3".
  • new argument in function ml_estimation(): wts for case weights.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("weibulltools")

1.0.1 by Hensel Tim-Gunnar, 8 months ago


Browse source code at https://github.com/cran/weibulltools


Authors: Hensel Tim-Gunnar [aut, cre]


Documentation:   PDF Manual  


GPL-2 license


Imports dplyr, LearnBayes, magrittr, plotly, Rcpp, sandwich, segmented, SPREDA, survival

Suggests ggplot2, knitr, rmarkdown, tidyverse

Linking to Rcpp, RcppArmadillo


See at CRAN