Response to comments on 'Log transformation: application and interpretation in biomedical research'

Modelling of intensive care unit (ICU) length of stay as a quality measure: a problematic exercise

BACKGROUND

Intensive care unit (ICU) length of stay (LOS) and the risk adjusted equivalent (RALOS) have been used as quality metrics. The latter measures entail either ratio or difference formulations or ICU random effects (RE), which have not been previously compared.

METHODS

From calendar year 2016 data of an adult ICU registry-database (Australia & New Zealand Intensive Care Society (ANZICS) CORE), LOS predictive models were established using linear (LMM) and generalised linear (GLMM) mixed models. Model fixed effects quality-metric formulations were estimated as RALOSR for LMM (geometric mean derived from log(ICU LOS)) and GLMM (day) and observed minus expected ICU LOS (OMELOS from GLMM). Metric confidence intervals (95%CI) were estimated by bootstrapping; random effects (RE) were predicted for LMM and GLMM. Forest-plot displays of ranked quality-metric point-estimates (95%CI) were generated for ICU hospital classifications (metropolitan, private, rural/regional, and tertiary). Robust rank confidence sets (point estimate and 95%CI), both marginal (pertaining to a singular ICU) and simultaneous (pertaining to all ICU differences), were established.

RESULTS

The ICU cohort was of 94,361 patients from 125 ICUs (metropolitan 16.9%, private 32.8%, rural/regional 6.4%, tertiary 43.8%). Age (mean, SD) was 61.7 (17.5) years; 58.3% were male; APACHE III severity-of-illness score 54.6 (25.7); ICU annual patient volume 1192 (702) and ICU LOS 3.2 (4.9). There was no concordance of ICU ranked model predictions, GLMM versus LMM, nor for the quality metrics used, RALOSR, OMELOS and site-specific RE for each of the ICU hospital classifications. Furthermore, there was no concordance between ICU ranking confidence sets, marginal and simultaneous for models or quality metrics.

CONCLUSIONS

Inference regarding adjusted ICU LOS was dependent upon the statistical estimator and the quality index used to quantify any LOS differences across ICUs. That is, there was no "one best model"; thus, ICU "performance" is determined by model choice and any rankings thereupon should be circumspect.

Enhancing precision in human neuroscience

Human neuroscience has always been pushing the boundary of what is measurable. During the last decade, concerns about statistical power and replicability - in science in general, but also specifically in human neuroscience - have fueled an extensive debate. One important insight from this discourse is the need for larger samples, which naturally increases statistical power. An alternative is to increase the precision of measurements, which is the focus of this review. This option is often overlooked, even though statistical power benefits from increasing precision as much as from increasing sample size. Nonetheless, precision has always been at the heart of good scientific practice in human neuroscience, with researchers relying on lab traditions or rules of thumb to ensure sufficient precision for their studies. In this review, we encourage a more systematic approach to precision. We start by introducing measurement precision and its importance for well-powered studies in human neuroscience. Then, determinants for precision in a range of neuroscientific methods (MRI, M/EEG, EDA, Eye-Tracking, and Endocrinology) are elaborated. We end by discussing how a more systematic evaluation of precision and the application of respective insights can lead to an increase in reproducibility in human neuroscience.

Single-cell RNA-seq data clustering: A survey with performance comparison study

Clustering analysis has been widely applied to single-cell RNA-sequencing (scRNA-seq) data to discover cell types and cell states. Algorithms developed in recent years have greatly helped the understanding of cellular heterogeneity and the underlying mechanisms of biological processes. However, these algorithms often use different techniques, were evaluated on different datasets and compared with some of their counterparts usually using different performance metrics. Consequently, there lacks an accurate and complete picture of their merits and demerits, which makes it difficult for users to select proper algorithms for analyzing their data. To fill this gap, we first do a review on the major existing scRNA-seq data clustering methods, and then conduct a comprehensive performance comparison among them from multiple perspectives. We consider 13 state of the art scRNA-seq data clustering algorithms, and collect 12 publicly available real scRNA-seq datasets from the existing works to evaluate and compare these algorithms. Our comparative study shows that the existing methods are very diverse in performance. Even the top-performance algorithms do not perform well on all datasets, especially those with complex structures. This suggests that further research is required to explore more stable, accurate, and efficient clustering algorithms for scRNA-seq data.

Explorations in statistics: the log transformation

Learning about statistics is a lot like learning about science: the learning is more meaningful if you can actively explore. This thirteenth installment of Explorations in Statistics explores the log transformation, an established technique that rescales the actual observations from an experiment so that the assumptions of some statistical analysis are better met. A general assumption in statistics is that the variability of some response Y is homogeneous across groups or across some predictor variable X. If the variability-the standard deviation-varies in rough proportion to the mean value of Y, a log transformation can equalize the standard deviations. Moreover, if the actual observations from an experiment conform to a skewed distribution, then a log transformation can make the theoretical distribution of the sample mean more consistent with a normal distribution. This is important: the results of a one-sample t test are meaningful only if the theoretical distribution of the sample mean is roughly normal. If we log-transform our observations, then we want to confirm the transformation was useful. We can do this if we use the Box-Cox method, if we bootstrap the sample mean and the statistic t itself, and if we assess the residual plots from the statistical model of the actual and transformed sample observations.

Delta method and bootstrap in linear mixed models to estimate a proportion when no event is observed: application to intralesional resection in bone tumor surgery

Resecting bone tumors requires good cutting accuracy to reduce the occurrence of local recurrence. This issue is considerably reduced with a navigated technology. The estimation of extreme proportions is challenging especially with small or moderate sample sizes. When no success is observed, the commonly used binomial proportion confidence interval is not suitable while the rule of three provides a simple solution. Unfortunately, these approaches are unable to differentiate between different unobserved events. Different delta methods and bootstrap procedures are compared in univariate and linear mixed models with simulations and real data by assuming the normality. The delta method on the z-score and parametric bootstrap provide similar results but the delta method requires the estimation of the covariance matrix of the estimates. In mixed models, the observed Fisher information matrix with unbounded variance components should be preferred. The parametric bootstrap, easier to apply, outperforms the delta method for larger sample sizes but it may be time costly. Copyright © 2016 John Wiley & Sons, Ltd.

Within-person variability in calculated risk factors: comparing the aetiological association of adiposity ratios with risk of coronary heart disease

BACKGROUND

Within-person variability in measured values of a risk factor can bias its association with disease. We investigated the extent of regression dilution bias in calculated variables and its implications for comparing the aetiological associations of risk factors.

METHODS

Using a numerical illustration and repeats from 42,300 individuals (12 cohorts), we estimated regression dilution ratios (RDRs) in calculated risk factors [body-mass index (BMI), waist-to-hip ratio (WHR), and waist-to-height ratio (WHtR)] and in their components (height, weight, waist circumference, and hip circumference), assuming the long-term average exposure to be of interest. Error-corrected hazard ratios (HRs) for risk of coronary heart disease (CHD) were compared across adiposity measures per standard-deviation (SD) change in: (i) baseline and (ii) error-corrected levels.

RESULTS

RDRs in calculated risk factors depend strongly on the RDRs, correlation, and comparative distributions of the components of these risk factors. For measures of adiposity, the RDR was lower for WHR [RDR: 0.72 (95% confidence interval 0.65-0.80)] than for either of its components [waist circumference: 0.87 (0.85-0.90); hip circumference: 0.90 (0.86-0.93) or for BMI: 0.96 (0.93-0.98) and WHtR: 0.87 (0.85-0.90)], predominantly because of the stronger correlation and more similar distributions observed between waist circumference and hip circumference than between height and weight or between waist circumference and height. Error-corrected HRs for BMI, waist circumference, WHR, and WHtR, were respectively 1.24, 1.30, 1.44, and 1.32 per SD change in baseline levels of these variables, and 1.24, 1.27, 1.35, and 1.30 per SD change in error-corrected levels.

CONCLUSIONS

The extent of within-person variability relative to between-person variability in calculated risk factors can be considerably larger (or smaller) than in its components. Aetiological associations of risk factors should be compared through the use of error-corrected HRs per SD change in error-corrected levels of these risk factors.