20 March 2018 in Statistical Modelling
Quantile regression quantifies the association of explanatory variables with a conditional quantile of a dependent variable without assuming any specific conditional distribution. It hence models the quantiles, instead of the mean as done in standard regression. In cases where either the requirements for mean regression, such as homoscedasticity, are violated or interest lies in the outer regions of the conditional distribution, quantile regression can explain dependencies more accurately than classical methods. However, many quantile regression papers are rather theoretical so the method has still not become a standard tool in applications. In this article, we explain quantile regression from an applied perspective. In particular, we illustrate the concept, advantages and disadvantages of quantile regression using two datasets as examples.
1 May 2016 in Methods of Information in Medicine
Summary: Background: Penalization and regularization techniques for statistical modeling have attracted increasing attention in biomedical research due to their advantages in the presence of high-dimensional data. A special focus lies on algorithms that incorporate automatic variable selection like the least absolute shrinkage operator (lasso) or statistical boosting techniques. Objectives: Focusing on the linear regression framework, this article compares the two most-common techniques for this task, the lasso and gradient boosting, both from a methodological and a practical perspective. Methods: We describe these methods highlighting under which circumstances their results will coincide in low-dimensional settings. In addition, we carry out extensive simulation studies comparing the performance in settings with more predictors than observations and investigate multiple combinations of noise-to-signal ratio and number of true non-zero coeffcients. Finally, we examine the impact of different tuning methods on the results. Results: Both methods carry out penalization and variable selection for possibly highdimensional data, often resulting in very similar models. An advantage of the lasso is its faster run-time, a strength of the boosting concept is its modular nature, making it easy to extend to other regression settings. Conclusions: Although following different strategies with respect to optimization and regularization, both methods imply similar constraints to the estimation problem leading to a comparable performance regarding prediction accuracy and variable selection in practice.
1 May 2016 in Journal of Cystic Fibrosis
To better understand the relative effects of infection with nontuberculous mycobacteria and Gram negative bacteria on lung function decline in cystic fibrosis, we assessed the impact of each infection in a Danish setting. Longitudinal registry study of 432 patients with cystic fibrosis contributing 53,771 lung function measures between 1974 and 2014. We used a mixed effects model with longitudinally structured correlation, while adjusting for clinically important covariates. Infections with a significant impact on rate of decline in %FEV1 were Mycobacterium abscessus complex with -2.22% points per year (95% CI -3.21 to -1.23), Burkholderia cepacia complex -1.95% (95% CI -2.51 to -1.39), Achromobacterxylosoxidans -1.55% (95% CI -2.21 to -0.90), and Pseudomonas aeruginosa -0.95% (95% CI -1.24 to -0.66). Clearing M. abscessus complex was associated with a change to a slower decline, similar in magnitude to the pre-infection slope. In a national population we have demonstrated the impact on lung function of each chronic CF pathogen. M. abscessus complex was associated with the worst impact on lung function. Eradication of M. abscessus complex may significantly improve lung function.
12 June 2013 in Statistical Modelling
Quantile regression provides a convenient framework for analyzing the impact of covariates on the complete conditional distribution of a response variable instead of only the mean. While frequentist treatments of quantile regression are typically completely nonparametric, a Bayesian formulation relies on assuming the asymmetric Laplace distribution as auxiliary error distribution that yields posterior modes equivalent to frequentist estimates. In this paper, we utilize a location-scale mixture of normals representation of the asymmetric Laplace distribution to transfer different flexible modelling concepts from Gaussian mean regression to Bayesian semiparametric quantile regression. In particular, we will consider high-dimensional geoadditive models comprising LASSO regularization priors and mixed models with potentially non-normal random effects distribution modeled via a Dirichlet process mixture. These extensions are illustrated using two large-scale applications on net rents in Munich and longitudinal measurements on obesity among children. The impact of the likelihood misspecification that underlies the Bayesian formulation of quantile regression is studied in terms of simulations.
12 September 2018 in BMC Cancer
Physical exercise and nutritional treatment are promising measures to prevent muscle wasting that is frequently observed in advanced-stage cancer patients. However, conventional exercise is not always suitable for these patients due to physical weakness and therapeutic side effects. In this pilot study, we examined the effect of a combined approach of the novel training method whole-body electromyostimulation (WB-EMS) and individualized nutritional support on body composition with primary focus on skeletal muscle mass in advanced cancer patients under oncological treatment. In a non-randomized controlled trial design patients (56.5% male; 59.9 ± 12.7 years) with advanced solid tumors (UICC III/IV, N = 131) undergoing anti-cancer therapy were allocated to a usual care control group (n = 35) receiving individualized nutritional support or to an intervention group (n = 96) that additionally performed a supervised physical exercise program in form of 20 min WB-EMS sessions (bipolar, 85 Hz) 2×/week for 12 weeks. The primary outcome of skeletal muscle mass and secondary outcomes of body composition, body weight and hand grip strength were measured at baseline, in weeks 4, 8 and 12 by bioelectrical impedance analysis and hand dynamometer. Effects of WB-EMS were estimated by linear mixed models. Secondary outcomes of physical function, hematological and blood chemistry parameters, quality of life and fatigue were assessed at baseline and week 12. Changes were analyzed by t-tests, Wilcoxon signed-rank or Mann-Whitney-U-tests. Twenty-four patients of the control and 58 of the WB-EMS group completed the 12-week trial. Patients of the WB-EMS group had a significantly higher skeletal muscle mass (0.53 kg [0.08, 0.98]; p = 0.022) and body weight (1.02 kg [0.05, 1.98]; p = 0.039) compared to controls at the end of intervention. WB-EMS also significantly improved physical function and performance status (p < 0.05). No significant differences of changes in quality of life, fatigue and blood parameters were detected between the study groups after 12 weeks. Supervised WB-EMS training is a safe strength training method and combined with nutritional support it shows promising effects against muscle wasting and on physical function in advanced-stage cancer patients undergoing treatment. ClinicalTrials.gov NCT02293239 (Date: November 18, 2014).
29 March 2017 in BMC Nephrology
Different measures of non-adherence to immunosuppressant (IS) medication have been found to be associated with rejection episodes after successful transplantation. The aim of the current study was to investigate whether graft rejection after renal transplantation is associated with patient-reported IS medication non-adherence and IS trough level variables (IS trough level variability and percentage of sub-therapeutic IS trough levels). Patient-reported non-adherence, IS trough level variability, percentage of sub-therapeutic IS trough levels, and acute biopsy-proven late allograft rejections were assessed in 267 adult renal transplant recipients who were ≥12 months post-transplantation. The rate of rejection was 13.5%. IS trough level variability, percentage of sub-therapeutic IS trough levels as well as patient-reported non-adherence were all significantly and positively associated with rejection, but not with each other. Logistic regression analyses revealed that only the percentage of sub-therapeutic IS trough levels and age at transplantation remained significantly associated with rejection. Particularly, the percentage of sub-therapeutic IS trough levels is associated with acute rejections after kidney transplantation whereas IS trough level variability and patient-reported non-adherence seem to be of subordinate importance. Patient-reported non-adherence and IS trough level variables were not correlated; thus, non-adherence should always be measured in a multi-methodological approach. Further research concerning the best combination of non-adherence measures is needed.