Matching on Generalized Propensity Scores with Continuous Exposures

In the context of a binary treatment, matching is a well-established approach in causal inference. However, in the context of a continuous treatment or exposure, matching is still underdeveloped. We propose an innovative matching approach to estimate an average causal exposure-response function under the setting of continuous exposures that relies on the generalized propensity score (GPS). Our approach maintains the following attractive features of matching: a) clear separation between the design and the analysis; b) robustness to model misspecification or to the presence of extreme values of the estimated GPS; c) straightforward assessments of covariate balance. We first introduce an assumption of identifiability, called local weak unconfoundedness. Under this assumption and mild smoothness conditions, we provide theoretical guarantees that our proposed matching estimator attains point-wise consistency and asymptotic normality. In simulations, our proposed matching approach outperforms existing methods under settings with model misspecification or in the presence of extreme values of the estimated GPS. We apply our proposed method to estimate the average causal exposure-response function between long-term PM2.5 exposure and all-cause mortality among 68.5 million Medicare enrollees, 2000-2016. We found strong evidence of a harmful effect of long-term PM2.5 exposure on mortality. Code for the proposed matching approach is provided in the CausalGPS R package, which is available on CRAN and provides a computationally efficient implementation.

The impact of climate change on children's nutritional status in coastal Bangladesh

This paper studies the impact of climate change on the nutritional status of very young children between the ages of 0-3 years by using weather data from the last half century merged with rich information on child, mother, and household characteristics in rural coastal Bangladesh. We evaluate the health consequences of rising temperature and relative humidity and varying rainfall jointly employing alternate functional forms. Leveraging models that control for annual trends and location-specific seasonality, and that allow the impacts of temperature to vary non-parametrically while rainfall and humidity have flexible non-linear forms, we find that temperatures that exceed 25 °C (the "comfortable" benchmark) in the month of birth exert negative effects on children's nutritional status as measured by mid upper arm circumference. Humidity has a positive impact which persists when child, mother and household controls are included. We find that exposure to changing climate in utero also matters. Explanations for these results include consequences of weather fluctuations on the extent of pasture, cropland, and rainfed lands planted with rice and other crops, and on mother's age at first marriage. Our results underline that climate change has real consequences for the health of very young populations in vulnerable areas.

Wilcoxon-Mann-Whitney odds ratio: A statistical measure for ordinal outcomes such as EDSS

BACKGROUND

In many clinical situations, ordinal scales afford the primary method of semi-quantifying patient outcomes. In the field of multiple sclerosis, the primary ordinal scale is the Expanded Disability Status Scale. Predominant methods of ordinal scale statistical analysis provide a p-value without effect size or rely heavily on the assumption of proportionality of odds, subjecting them to lack of power and error. The Wilcoxon-Manny-Whitney Odds is a statistical method which provides significant information such as p-value, effect size, number needed to treat, confidence intervals, and is largely assumption-free. However, its utility has not been demonstrated in the field of multiple sclerosis.

METHODS

Three clinical studies in the field of multiple sclerosis were selected which utilized ordinal scale outcomes at group or individual levels. Data from these studies was extracted using WebPlotDigitizer, and a custom Wilxocon-Mann-Whitney Odds software was applied to each dataset to re-analyze the main outcomes of the studies.

RESULTS

Re-analysis of the manuscript by Muraro et al., 2017 demonstrated that autologous stem cell transplantation for relapsing remitting multiple sclerosis resulted in a 65% chance of improving from any Expanded Disability Status Scale category, although not significant. Re-analysis of the manuscript by Songthammawat et al., 2019 demonstrated chance of improvement with intravenous methylprednisolone and concurrent plasma exchange was 185% versus 32% in intravenous methylprednisolone with add-on plasma exchange, although not significant. Re-analysis of Kister et al., 2012 demonstrated the chances of mobility or cognition scores generally favored decline at every 5-year increment of study, and although statistically significant, these were smaller effect sizes ranging from an 11% chance of improvement to a 66% chance of decline over a 5-year interval.

DISCUSSION

The Wilcoxon-Mann-Whitney Odds simplifies ordinal data analysis with its robust largely assumption-free nature. In the place of numerous statistical tests, this single test provides effect size estimate, number needed to treat, p-values, and confidence intervals. Importantly, the Wilcoxon-Mann-Whitney Odds effect size calculation is intuitively applicable to both individual and population-levels. Further, the Wilcoxon-Mann-Whitney Odds allows intuitive description of the progression of large cohorts over time, and we were able to clearly convey the odds of mobility and cognitive decline over 30 years in a large multiple sclerosis cohort. Overall, the Wilcoxon-Mann-Whitney Odds is a powerful and robust statistical test with significant promise within the field of multiple sclerosis.

New confinement index and new perspective for comparing countries - COVID-19

BACKGROUND AND OBJECTIVE

In the difficult problem of comparing countries regarding their lockdown measures or deaths caused by the COVID-19, there is still no agreement on what is the best strategy to follow. Thus, we propose a new way of comparison countries that avoids the main difficulties in the comparison by using three-dimensional trajectories for this type of data.

METHODS

We introduce a new index to analyze the level of confinement that each country was subject to overtime, based on the Community Mobility Reports published by Google resorting to Principal Component Analysis. Subsequently, by using longitudinal clustering, we divide the European countries into similar groups according to the COVID-19 obits and also to the confinement index. However, to make the most out of the clustering methods we resort to artificial longitudinal data to evaluate both the methods and the indices.

RESULTS

By using artificial data, we discover that Calinski-Harabasz outperformed other internal indices in indicating the real number of clusters. The tests also suggested that K-means with Euclidean distance was the best method among the ones studied. With the application to both the mobility and fatalities datasets, we found two groups in each one.

CONCLUSIONS

Our analysis enables us to discover that European northern countries had more mobility during the first confinement and that the deaths caused by COVID-19 started to drop around the 40th day since the first death.

The impact of public-private partnerships Investment in Energy on carbon emissions: evidence from nonparametric causality-in-quantiles

The current investigation examines causal relation between PPP investment in energy sector and CO2 emissions in selected developing countries by using non-parametric causality in quantiles and linear granger causality techniques. Range of the data is from January 1998 to December 2016. Although results obtained by linear granger causality test does not report any causal relation between PPP investment in energy and CO2 emissions, but findings from non-parametric test show that non-linear relationship exhibit between variables. The non-parametric outcomes indicate that PPP investment in non-renewable energy in the selected countries contribute to carbon emissions and thus degrade environment. And therefore, the need is to divert PPP investment to renewable energy where it is more effective. This investigation provides valuable information to policy-makers in developing countries to think out of box and address pressing environmental issues.

Yield stability analysis of orange - Fleshed sweet potato in Indonesia using AMMI and GGE biplot

Orange-Fleshed Sweet Potato (OFSP) is an important crop in Indonesia. Yield potential and genotypic adaptability are important factors in varietal development. The purpose of this study was to estimate the stability of yield and to select the best OFSP genotypes across three agroecosystems in West Java, Indonesia. The field trials used were augmented design with 50 F1 Orange-Fleshed Sweet Potato (OFSP) genotypes as treatment, and seven check varieties as controls. The experiments were conducted in three different agroecosystems in West Java (Sumedang, Bandung, and Karawang). Selection was based on physical characteristics of sweet potato tuber, yield and stability across three environments. Data analysis of the yield characters, yield component, and tuber quality were performed by combined variance analysis. Selected genotypes were analyzed for stability yield using the parametric, non-parametric, Additive Main effects and Multiplicative Interaction (AMMI), AMMI Stability Value (ASV), and Genotype and Genotype by Environment (GGE) biplot models. Results identified the top best ten F1 genotypes namely F1-38 (G1), F1-69 (G2), F1-71 (G3), F1-77 (G4), F1-127 (G5), F1-128 (G6), F1-135 (G7), F1-159 (G8), F1-191 (G9), and F1-226 (G10). Location showed a significant effect on yield. Genotypes F1-069, F1-077, F1-226, F1-038, and F1-128 have the lowest ASR based on non-parametric and parametric stability models and there were identified as the most stable. AMMI analysis identified F1-128, F1-135, F1-038, and F1-069 as the most stable genotypes. F1-38 (G1), F1-69 (G2), F1-128 (G6) were found to be the most stable genotypes based on ASV analysis, while GGE biplot identified F1-38 (G1) and F1-69 (G2) genotypes as the stable genotypes. Other genotypes were considered to as location-specific. Based on AMMI, ASV, and GGE Biplot models, F1-038, and F1-069 were identified as stable genotypes. They produced higher yields than other genotypes. Therefore, the F1-038 and F1-069 genotypes can be potentially recommended as superior varieties for West Java, Indonesia.

A novel method for interpreting survival analysis data: description and test on three major clinical trials on cardiovascular prevention

Background

Major results of randomized clinical trials on cardiovascular prevention are currently provided in terms of relative or absolute risk reductions, including also the number needed to treat (NNT), incorrectly implying that a treatment might prevent the occurrence of the outcome/s under investigation. Provided that these results are based on survival analysis, the primary measure of which is time-to-the outcome and not the outcome itself, we sought an alternative method to describe, analyse and interpret clinical trial results consistent with this assumption, so as to better define qualitative and quantitative heterogeneity of various therapeutic strategies in terms of their effects and costs.

Methods

The original Kaplan-Meier graphs of three major positive cardiovascular prevention trials (PROVE-IT, LIFE and HOPE) were captured from the PDF images of the article and then digitalized. We calculated the difference between the placebo and active treatment curves and plotted it as a function of time to describe the event-free time gain (Time-Gain) produced by the active treatment. By calculating the exposure to the active treatment in terms of months (MoT) as a function of time and dividing it for the corresponding time-dependent number of event-free years gained (i.e. months/12), we described the kinetics of the pharmaco-economic index MoT/y⁺. The same procedure was repeated replacing MoT with the actual number of patients being treated at each time point as a function of time to obtain the NNT to gain 1 event-free year (NNT/y⁺) curve.

Results

The Time-Gain curves depict the kinetics of the treatment-related effect over time and possess the peculiar feature of being smooth and accurately fitted by second-order polynomial functions (a*time² + b*time); similarly, also the MoT/y⁺ and NNT/y⁺ curves can be accurately fitted by power functions (a*time^b).

These curves and indices allow to fully appreciate the quantitative and qualitative heterogeneity, both in terms of effects and costs, of the different therapeutic strategies adopted in the three trials.

Conclusions

With our novel method, by exploiting original Kaplan-Meier curves from three major clinical trials on cardiovascular prevention, we generate new information on the actual consequences of choosing a therapeutic strategy vs another, thus ultimately providing the clinical gain in terms of time-dependent functions. Accurately assessing clinically and economic meaningful results from any intervention trial reporting positive results through this approach, facilitates objective comparisons and increases reliability in predicting survival among the various therapeutic options provided.

Trial registration

PROVE-IT (Pravastatin or Atorvastatin Evaluation and Infection Therapy (TIMI22), Clinical trial registration number: NCT00382460, date of registration: September 29, 2006, study start date: November 2000).

LIFE (Losartan Intervention For Endpoint Reduction in Hypertension (LIFE) Study, Clinical trial registration number: NCT00338260, date of registration: June 20, 2006, study start date: June 1995).

HOPE (Heart Outcomes Prevention Evaluation; we could not find Clinical trial registration number and date of registration).

The composition characteristics of different crop straw types and their multivariate analysis and comparison

The heterogeneity and complex composition of crop straw are some of the main obstacles to its scientific and efficient industrial utilization. To thoroughly reveal and identify the composition of different crop straw types and their latent attributes, in this study, 784 straw samples of rice, wheat, corn, rape and cotton were collected. Based on the large sample size, 18 composition characteristics, including chemical composition, proximate composition, ultimate composition, and heating values, were adopted to determine the profiles of the crop straw composition characteristics. Correlation analysis and 7 different types of multivariate analysis were applied and compared. The results indicated that among the 18 characteristics, hemicellulose, water-soluble carbohydrates, crude proteins, phosphorus, fixed carbon, hydrogen, nitrogen, and sulfur had non-normal distributions. Spearman method was a more suitable correlation analysis approach for the crop straw characteristics than Pearson method. The results of the different multivariate analysis methods were reflected in the different classification attributes of water-soluble carbohydrates, phosphorus, hydrogen and sulfur. Non-parametric principal component analysis and non-parametric exploratory factor analysis provided consistent results. The characteristics could be divided into 4 categories of intrinsic associated attributes, namely, (1) lignin, volatile matter, fixed carbon, carbon, hydrogen, higher heating value, and lower heating value; (2) potassium, ash, and sulfur; (3) cellulose, hemicellulose, moisture, and oxygen; and (4) water-soluble carbohydrates, crude proteins, phosphorus, and nitrogen, which exhibited combustion positive, combustion negative, biochemical conversion, and nutritional property, respectively. The study results provide data and methodology support for the development of crop straw utilization strategies.

Non-parametric and semi-parametric support estimation using SEquential RESampling random walks on biomolecular sequences

Non-parametric and semi-parametric resampling procedures are widely used to perform support estimation in computational biology and bioinformatics. Among the most widely used methods in this class is the standard bootstrap method, which consists of random sampling with replacement. While not requiring assumptions about any particular parametric model for resampling purposes, the bootstrap and related techniques assume that sites are independent and identically distributed (i.i.d.). The i.i.d. assumption can be an over-simplification for many problems in computational biology and bioinformatics. In particular, sequential dependence within biomolecular sequences is often an essential biological feature due to biochemical function, evolutionary processes such as recombination, and other factors. To relax the simplifying i.i.d. assumption, we propose a new non-parametric/semi-parametric sequential resampling technique that generalizes “Heads-or-Tails” mirrored inputs, a simple but clever technique due to Landan and Graur. The generalized procedure takes the form of random walks along either aligned or unaligned biomolecular sequences. We refer to our new method as the SERES (or “SEquential RESampling”) method. To demonstrate the performance of the new technique, we apply SERES to estimate support for the multiple sequence alignment problem. Using simulated and empirical data, we show that SERES-based support estimation yields comparable or typically better performance compared to state-of-the-art methods.

Farmers' choice of genotypes and trait preferences in tropically adapted chickens in five agro-ecological zones in Nigeria

This study aimed at determining chicken genotypes of choice and traits preference in chicken by smallholder farmers in Nigeria. Data were obtained from a total of 2063 farmers using structured questionnaires in five agro-ecological zones in Nigeria. Chi square (χ²) statistics was used to explore relationships between categorical variables. The mean ranks of the six genotypes and twelve traits of preference were compared using the non-parametric Kruskal-Wallis H (with Mann-Whitney U test for post hoc separation of mean ranks), Friedman, and Wilcoxon signed-rank (with Bonferroni's adjustments) tests. Categorical principal component analysis (CATPCA) was used to assign farmers into groups. Gender distribution of farmers was found to be statistically significant (χ² = 16.599; P ≤ 0.002) across the zones. With the exception of Shika Brown, preferences for chicken genotypes were significantly (P ≤ 0.01) influenced by agro-ecological zone. However, gender differentiated response was only significant (P ≤ 0.01) in Sasso chicken with more preference by male farmers. Overall, FUNAAB Alpha, Sasso, and Noiler chicken were ranked 1st, followed by Kuroiler (4th), Shika Brown (5th), and Fulani birds (6th), respectively. Within genotypes, within and across zones and gender, preferences for traits varied significantly (P ≤ 0.005 and P ≤ 0.01). Traits of preference for selection of chicken breeding stock tended towards body size, egg number, egg size, and meat taste. Spearman's rank order correlation coefficients of traits of preference were significant (P ≤ 0.01) and ranged from 0.22 to 0.90. The two PCs extracted, which explained 65.3% of the variability in the dataset, were able to assign the farmers into two groups based on preference for body size of cock and hen and the other ten traits combined. The present findings may guide the choice of appropriate chicken genotypes while the traits of economic importance may be incorporated into future genetic improvement and conservation programs in Nigeria.

Semi-supervised neighborhoods and localized patient outcome prediction

Robust statistical methods that can provide patients and their healthcare providers with individual predictions are needed to help guide informed medical decisions. Ideally an individual prediction would display the full range of possible outcomes (full predictive distribution), would be obtained with a user-specified level of precision, and would be minimally reliant on statistical model assumptions. We propose a novel method that satisfies each of these criteria via the semi-supervised creation of an axis-parallel covariate neighborhood constructed around a given point defining the patient of interest. We then provide non-parametric estimates of the outcome distribution for the subset of subjects in this neighborhood, which we refer to as a localized prediction. We implement local prediction methods using dynamic graphical methods which allow the user to vary key options such as the choice of the variables defining the neighborhood, and the size of the neighborhood.

GSAR: Bioconductor package for Gene Set analysis in R

BACKGROUND

Gene set analysis (in a form of functionally related genes or pathways) has become the method of choice for analyzing omics data in general and gene expression data in particular. There are many statistical methods that either summarize gene-level statistics for a gene set or apply a multivariate statistic that accounts for intergene correlations. Most available methods detect complex departures from the null hypothesis but lack the ability to identify the specific alternative hypothesis that rejects the null.

RESULTS

GSAR (Gene Set Analysis in R) is an open-source R/Bioconductor software package for gene set analysis (GSA). It implements self-contained multivariate non-parametric statistical methods testing a complex null hypothesis against specific alternatives, such as differences in mean (shift), variance (scale), or net correlation structure. The package also provides a graphical visualization tool, based on the union of two minimum spanning trees, for correlation networks to examine the change in the correlation structures of a gene set between two conditions and highlight influential genes (hubs).

CONCLUSIONS

Package GSAR provides a set of multivariate non-parametric statistical methods that test a complex null hypothesis against specific alternatives. The methods in package GSAR are applicable to any type of omics data that can be represented in a matrix format. The package, with detailed instructions and examples, is freely available under the GPL (> = 2) license from the Bioconductor web site.

Population pharmacokinetics of nevirapine in Malaysian HIV patients: a non-parametric approach

AIMS

Nevirapine is the first non-nucleoside reverse-transcriptase inhibitor approved and is widely used in combination therapy to treat HIV-1 infection. The pharmacokinetics of nevirapine was extensively studied in various populations with a parametric approach. Hence, this study was aimed to determine population pharmacokinetic parameters in Malaysian HIV-infected patients with a non-parametric approach which allows detection of outliers or non-normal distribution contrary to the parametric approach.

METHODS

Nevirapine population pharmacokinetics was modelled with Pmetrics. A total of 708 observations from 112 patients were included in the model building and validation analysis. Evaluation of the model was based on a visual inspection of observed versus predicted (population and individual) concentrations and plots weighted residual error versus concentrations. Accuracy and robustness of the model were evaluated by visual predictive check (VPC). The median parameters' estimates obtained from the final model were used to predict individual nevirapine plasma area-under-curve (AUC) in the validation dataset. The Bland-Altman plot was used to compare the AUC predicted with trapezoidal AUC.

RESULTS

The median nevirapine clearance was of 2.92 L/h, the median rate of absorption was 2.55/h and the volume of distribution was 78.23 L. Nevirapine pharmacokinetics were best described by one-compartmental with first-order absorption model and a lag-time. Weighted residuals for the model selected were homogenously distributed over the concentration and time range. The developed model adequately estimated AUC.

CONCLUSIONS

In conclusion, a model to describe the pharmacokinetics of nevirapine was developed. The developed model adequately describes nevirapine population pharmacokinetics in HIV-infected patients in Malaysia.

Transition probability estimates for non-Markov multi-state models

Non-parametric estimation of the transition probabilities in multi-state models is considered for non-Markov processes. Firstly, a generalization of the estimator of Pepe et al., (1991) (Statistics in Medicine) is given for a class of progressive multi-state models based on the difference between Kaplan-Meier estimators. Secondly, a general estimator for progressive or non-progressive models is proposed based upon constructed univariate survival or competing risks processes which retain the Markov property. The properties of the estimators and their associated standard errors are investigated through simulation. The estimators are demonstrated on datasets relating to survival and recurrence in patients with colon cancer and prothrombin levels in liver cirrhosis patients.

Parameter-less approaches for interpreting dynamic cellular response

Cellular response such as cell signaling is an integral part of information processing in biology. Upon receptor stimulation, numerous intracellular molecules are invoked to trigger the transcription of genes for specific biological purposes, such as growth, differentiation, apoptosis or immune response. How complex are such specialized and sophisticated machinery? Computational modeling is an important tool for investigating dynamic cellular behaviors. Here, I focus on certain types of key signaling pathways that can be interpreted well using simple physical rules based on Boolean logic and linear superposition of response terms. From the examples shown, it is conceivable that for small-scale network modeling, reaction topology, rather than parameter values, is crucial for understanding population-wide cellular behaviors. For large-scale response, non-parametric statistical approaches have proven valuable for revealing emergent properties.

The use and misuse of statistical methodologies in pharmacology research

Descriptive, exploratory, and inferential statistics are necessary components of hypothesis-driven biomedical research. Despite the ubiquitous need for these tools, the emphasis on statistical methods in pharmacology has become dominated by inferential methods often chosen more by the availability of user-friendly software than by any understanding of the data set or the critical assumptions of the statistical tests. Such frank misuse of statistical methodology and the quest to reach the mystical α<0.05 criteria has hampered research via the publication of incorrect analysis driven by rudimentary statistical training. Perhaps more critically, a poor understanding of statistical tools limits the conclusions that may be drawn from a study by divorcing the investigator from their own data. The net result is a decrease in quality and confidence in research findings, fueling recent controversies over the reproducibility of high profile findings and effects that appear to diminish over time. The recent development of "omics" approaches leading to the production of massive higher dimensional data sets has amplified these issues making it clear that new approaches are needed to appropriately and effectively mine this type of data. Unfortunately, statistical education in the field has not kept pace. This commentary provides a foundation for an intuitive understanding of statistics that fosters an exploratory approach and an appreciation for the assumptions of various statistical tests that hopefully will increase the correct use of statistics, the application of exploratory data analysis, and the use of statistical study design, with the goal of increasing reproducibility and confidence in the literature.