A Regression Framework for Rank Tests Based on the Probabilistic Index Model

Neve, Jan De; Thas, Olivier

doi:10.1080/01621459.2015.1016226

Cited by 25 publications

(32 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is also showed that the estimatorβ is asymptotically normally distributed with a consistent estimator of its variance. We refer readers to [25], [8] and the supplementary file for further details about PIMs.…”

Section: Probabilistic Index Modelsmentioning

confidence: 99%

“…We propose a semi-parametric method based on Probabilistic Index Models (PIM) [25,8], for testing DE in scRNA-seq data. PIMs entail a large class of semi-parametric models that can generate many of the traditional rank tests, such as the Wilcoxon-Mann-Whitney test and the Kruskal-Wallis test [25,8]. These models can be seen as the rank-equivalent of the generalized linear models (GLM).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Probabilistic index models for testing differential expression in single cell RNA sequencing data

Assefa¹,

Vandesompele²,

Thas³

2019

Preprint

Self Cite

View full text Add to dashboard Cite

Single-cell RNA sequencing (scRNA-seq) technologies profile gene expression patterns in individual cells. It is often of interest to test for differential expression (DE) between conditions, e.g. treatment and control or between cell types. Simulation studies have shown that non-parametric tests, such as the Wilcoxon-rank sum test, can robustly detect significant DE, with better performance than many parametric tools specifically developed for scRNA-seq data analysis. However, these classical rank tests cannot be used for complex experimental designs involving multiple groups, multiple factors and confounding variables. Further, rank based tests do not provide an interpretable measure of the effect size. We propose a semi-parametric approach based on probabilistic index models (PIM) that form a flexible class of models that generalize classical rank tests. Our method does not rely on strong distributional assumptions and it allows accounting for confounding factors. Moreover, our method allows for the estimation of the effect size in terms of a probabilistic index. Real data analysis demonstrated that PIM is capable of identifying biologically meaningful DE. Our simulation studies also show that tests for DE succeed well in controlling the false discovery rate at its nominal level, while maintaining good sensitivity as compared to competing methods.Keywords: differential expression, single-cell RNA-sequencing, probabilistic index models, semi-parametric. tool. We also explored the cross data agreement of all tools in detecting DE genes. In addition to DE testing, we performed gene set enrichment analysis (GSEA) using the fgsea[21] R bioconductor package (version 1.8.0). In particular, we used a set TP53 pathway genes obtained from a study by Fischer [10]. Genes were ranked based their test statistic as in [29].

show abstract

Section: Probabilistic Index Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Probabilistic index models for testing differential expression in single cell RNA sequencing data

Assefa¹,

Vandesompele²,

Thas³

2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Their interpretation is as follows: if p ij ⩽ p rs the observations under factor combination ( i , j ) tend to result in smaller values as the observations under factor combination ( r , s ). Moreover, another interpretation can be given in terms of additive effects by using a decomposition of the distribution functions as in Akritas and Arnold () and the supporting information in De Neve and Thas (): writing

G = {\overset{F}{true}}_{\cdot \cdot} = false(1 false/ a b false) \sum_{i = 1}^{a} \sum_{j = 1}^{b} F_{i j},

A_{i} = {\overset{F}{true}}_{i \cdot} - G = false(1 false/ b false) \sum_{j = 1}^{b} F_{i j} - G,

B_{j} = {\overset{F}{true}}_{\cdot j} - G = false(1 / a false) \sum_{i = 1}^{a} F_{i j} - G

and

{(A B)}_{i j} = F_{i j} - {\overset{F}{true}}_{i \cdot} - {\overset{F}{true}}_{\cdot j} + G

we have F ij = G + A i + B j +( AB ) ij . Plugging this into the definition of the non‐parametric effect results in an additive effects representation as in classical linear models

false0 p_{i j} = \int G normald F_{i j} = \frac{1}{2} + \int G d A_{i} + \int G d B_{j} + \int G d {(A B)}_{<...>}

…”

Section: Specific Designsmentioning

confidence: 99%

“…Recently, Fan and Zhang () have proposed a generalized estimating equations approach for rank‐transformed data and Thas et al . () and De Neve and Thas () have introduced the similar concept of the so‐called probabilistic index model (PIM). It allows flexible rank‐based modelling for various designs.…”

Section: Introductionmentioning

confidence: 99%

“…It allows flexible rank‐based modelling for various designs. These models, however, are based on weighted effects, where sample sizes are involved and the related inference procedures are mainly developed for null hypotheses formulated in terms of the distribution functions (with the exception of De Neve and Thas (), who considered null hypotheses in terms of probabilistic indices for special models such as the unpaired two‐sample design).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Rank-Based Procedures in Factorial Designs: Hypotheses About Non-Parametric Treatment Effects

Brunner

Konietschke

Pauly

et al. 2016

Journal of the Royal Statistical Society Series B: Statistical Methodology

123

View full text Add to dashboard Cite

Abstract. Existing tests for factorial designs in the nonparametric case are based on hypotheses formulated in terms of distribution functions. Typical null hypotheses, however, are formulated in terms of some parameters or effect measures, particularly in heteroscedastic settings. Here this idea is extended to nonparametric models by introducing a novel nonparametric ANOVA-type-statistic based on ranks which is suitable for testing hypotheses formulated in meaningful nonparametric treatment effects in general factorial designs. This is achieved by a careful in-depth study of the common distribution of rank-based estimators for the treatment effects. Since the statistic is asymptotically not a pivotal quantity we propose three different approximation techniques, discuss their theoretic properties and compare them in extensive simulations together with two additional Wald-type tests. An extension of the presented idea to general repeated measures designs is briefly outlined. The proposed rank-based procedures maintain the pre-assigned type-I error rate quite accurately, also in unbalanced and heteroscedastic models.All authors are given in alphabetic order:

show abstract

Association of Parental Status and Gender With Burden of Multidisciplinary Tumor Boards Among Oncology Physicians

Chau,

LaGuardia,

Kim

et al. 2023

JAMA Netw Open

View full text Add to dashboard Cite

ImportanceTumor boards are integral to the care of patients with cancer. However, data investigating the burden of tumor boards on physicians are limited.ObjectiveTo investigate what physician-related and tumor board–related factors are associated with higher tumor board burden among oncology physicians.Design, Setting, and ParticipantsTumor board burden was assessed by a cross-sectional convenience survey posted on social media and by email to Cedars-Sinai Medical Center cancer physicians between March 3 and April 3, 2022. Tumor board start times were independently collected by email from 22 top cancer centers.Main Outcomes and MeasuresTumor board burden was measured on a 4-point scale (1, not at all burdensome; 2, slightly burdensome; 3, moderately burdensome; and 4, very burdensome). Univariable and multivariable probabilistic index (PI) models were performed.ResultsSurveys were completed by 111 physicians (median age, 42 years [IQR, 36-50 years]; 58 women [52.3%]; 60 non-Hispanic White [54.1%]). On multivariable analysis, factors associated with higher probability of tumor board burden included radiology or pathology specialty (PI, 0.68; 95% CI, 0.54-0.79; P = .02), attending 3 or more hours per week of tumor boards (PI, 0.68; 95% CI, 0.58-0.76; P &lt; .001), and having 2 or more children (PI, 0.65; 95% CI, 0.52-0.77; P = .03). Early or late tumor boards (before 8 am or at 5 pm or after) were considered very burdensome by 33 respondents (29.7%). Parents frequently reported a negative burden on childcare (43 of 77 [55.8%]) and family dynamics (49 of 77 [63.6%]). On multivariable analysis, a higher level of burden from early or late tumor boards was independently associated with identifying as a woman (PI, 0.69; 95% CI, 0.57-0.78; P = .003) and having children (PI, 0.75; 95% CI, 0.62-0.84; P &lt; .001). Independent assessment of 358 tumor boards from 22 institutions revealed the most common start time was before 8 am (88 [24.6%]).Conclusions and RelevanceThis survey study of tumor board burden suggests that identifying as a woman or parent was independently associated with a higher level of burden from early or late tumor boards. The burden of early or late tumor boards on childcare and family dynamics was commonly reported by parents. Having 2 or more children, attending 3 or more hours per week of tumor boards, and radiology or pathology specialty were associated with a significantly higher tumor board burden overall. Future strategies should aim to decrease the disparate burden on parents and women.

show abstract

A Regression Framework for Rank Tests Based on the Probabilistic Index Model

Cited by 25 publications

References 30 publications

Probabilistic index models for testing differential expression in single cell RNA sequencing data

Probabilistic index models for testing differential expression in single cell RNA sequencing data

Rank-Based Procedures in Factorial Designs: Hypotheses About Non-Parametric Treatment Effects

Association of Parental Status and Gender With Burden of Multidisciplinary Tumor Boards Among Oncology Physicians

Contact Info

Product

Resources

About