bioRxiv Science⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/137232

Are Your Covariates Under Control? How Normalization Can Re-introduce Covariate Effects

Pain, O.·Dudbridge, F.·Ronald, A.
View original source ↗Export metadata ↓

Abstract

Many statistical tests rely on the assumption that the residuals of a model are normally distributed. Rank-based inverse normal transformation (INT) of the dependent variable is one of the most popular approaches to satisfy the normality assumption. Studies regularly adjust for covariates and then normalize the residuals. This study investigated the effect of regressing covariates against the dependent variable and then applying rank-based INT to the residuals. The correlation between the dependent variable and covariates at each stage of processing was assessed. An alternative approach was tested of applying rank-based INT to the dependent variable before regressing covariates was tested. Analyses based on both simulated and real data examples demonstrated that applying rank-based INT to the dependent variable residuals after regressing out covariates re-introduces a linear correlation between the dependent variable and covariates in almost all situations. This will increase type-1 errors and reduce power. Our proposed alternative approach, where rank-based INT was applied prior to controlling for covariate effects, gave residuals that were normally distributed and linearly uncorrelated with covariates. This approach is therefore recommended.

Source connections

Explore related subjects

epidemiologyPreprint

Keep this discovery

BibTeXRIS

Pain, O., Dudbridge, F., Ronald, A.. 2017-05-15. Are Your Covariates Under Control? How Normalization Can Re-introduce Covariate Effects. https://doi.org/10.1101/137232

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2018

Translating surveillance data into incidence estimates

Monitoring a population for a disease requires the hosts to be sampled and tested for the pathogen. This results in sampling series from which to estimate the disease incidence, i.e. the proportion of hosts infected. Existing estimation methods assume that disease incidence is not changing between monitoring rounds, resulting in underestimation of the disease incidence. In this paper we develop an incidence estimation model accounting for epidemic growth with monitoring rounds sampling varying incidence. We also show how to accommodate the asymptomatic period characteristic to most diseases. For practical use, we produce an approximation of the model, which is subsequently shown accurate for relevant epidemic and sampling parameters. Both the approximation and the full model are applied to stochastic spatial simulations of epidemics. The results prove their consistency for a very wide range of situations.

epidemiology↗
bioRxiv2018

The Swiss Primary Ciliary Dyskinesia registry: objectives, methods and first results

Primary Ciliary Dyskinesia (PCD) is a rare hereditary, multi-organ disease caused by defects in ciliary structure and function. It results in a wide range of clinical manifestations, most commonly in the upper and lower airways. Central data collection in national and international registries is essential to studying the epidemiology of rare diseases and filling in gaps in knowledge of diseases such as PCD. For this reason, the Swiss Primary Ciliary Dyskinesia Registry (CH-PCD) was founded in 2013 as a collaborative project between epidemiologists and adult and paediatric pulmonologists.\n\nThe registry records patients of any age, suffering from PCD, who are treated and resident in Switzerland. It collects information from patients identified through physicians, diagnostic facilities, and patient organisations. The registry dataset contains data on diagnostic evaluations, lung function, microbiology and imaging, symptoms, treatments, and hospitalizations.\n\nBy May 2018, CH-PCD has contacted 566 physicians of different specialties and identified 134 patients with PCD. At present this number represents an overall 1 in 63,000 prevalence of people diagnosed with PCD in Switzerland. Prevalence differs by age and region; it is highest in children and adults younger than 30 years, and in Espace Mittelland. The median age of patients in the registry is 25 years (range 5-73), and 49 patients have a definite PCD diagnosis based on recent international guidelines. Data from CH-PCD are contributed to international collaborative studies and the registry facilitates patient identification for nested studies.\n\nCH-PCD has proven to be a valuable research tool that already has highlighted weaknesses in PCD clinical practice in Switzerland. Development of centralised diagnostic and management centres and adherence to international guidelines are needed to improve diagnosis and management--particularly for adult PCD patients.

epidemiology↗
bioRxiv2018

Perfect Counterfactuals for Epidemic Simulations

Simulation studies are often used to predict the expected impact of control measures in infectious disease outbreaks. Typically, two independent sets of simulations are conducted, one with the intervetnion, and one without, and epidemic sizes (or some related metric) are compared to estimate the effect of the intervention. Since it is possible that controlled epidemics are larger than uncontrolled ones if there is substantial stochastic variation between epidemics, uncertainty intervals from this approach can include a negative effect even for an effective intervention. To more precisely estimate the number of cases an intervention will prevent within a single epidemic, here we develop a single world approach to matching simulations of controlled epidemics to their exact uncontrolled counterfac-tual. Our method borrows concepts from percolation approaches prune out possible epidemic histories and create potential epidemic graph that can be realized to create perfectly matched controlled and uncontrolled epidemics. We present an implementation of this method for a common class of compartmental models, and its application in a simple SIR model. Results illustrate how, at the cost of some computation time, this method substantially narrows confidence intervals and avoids non-sensical inferences.

epidemiology↗