An empirical comparison of two novel transformation models

Tian, Yuqi; Hothorn, Torsten; Li, Chun; Harrell, Frank E.; Shepherd, Bryan E.

doi:10.1002/sim.8425

Cited by 11 publications

(16 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We used a proportional odds logistic ordinal regression model for continuous data [i.e., a cumulative probability model with logit link (61,62)] to assess the impact of distinct regressors on multisensory spatial processing. For the CCE task, we regressed the mean difference in RT during congruent and incongruent visuo-tactile stimulation, as well as the change in accuracy, on gender, age, full-scale IQ, baseline tactile performance, and diagnostic group.…”

Section: Analyses and Statistical Modelingmentioning

confidence: 99%

Visual-Tactile Spatial Multisensory Interaction in Adults With Autism and Schizophrenia

et al. 2020

View full text Add to dashboard Cite

Background: Individuals with autism spectrum disorder (ASD) and schizophrenia (SZ) exhibit multisensory processing difficulties and social impairments, with growing evidence that the former contributes to the latter. However, this work has largely reported on separate cohorts, introducing method variance as a barrier to drawing broad conclusions across studies. Further, very few studies have addressed touch, resulting in sparse knowledge about how these two clinical groups may integrate somatic information with other senses.Methods: In this study, we compared adults with ASD (n = 29), SZ (n = 24), and typical developmental histories (TD, n = 37) on two tasks requiring visual-tactile spatial multisensory processing. In the first task (crossmodal congruency), participants judged the location of a tactile stimulus in the presence or absence of simultaneous visual input that was either spatially congruent or incongruent, with poorer performance for incongruence an index of spatial multisensory interaction. In the second task, participants reacted to touch in the presence or absence of dynamic visual stimuli that appeared to approach or recede from the body. Within a certain radius around the body, defined as peripersonal space (PPS), an approaching visual or auditory stimulus reliably speeds reaction times (RT) to touch; outside of this radius, in extrapersonal space (EPS), there is no multisensory effect. PPS can be defined both by its size (radius) and slope (sharpness of the PPS-EPS boundary). Clinical measures were administered to explore relations with visual-tactile processing.Results: Neither clinical group differed from controls on the crossmodal congruency task. The ASD group had significantly smaller and more sharply-defined PPSs compared to the other two groups. Small PPS size was related to social symptom severity across groups, but was largely driven by the TD group, without significant effects in either clinical group.Noel et al. Visual-Tactile in Autism and SchizophreniaConclusions: These results suggest that: (1) spatially static visual-tactile facilitation is intact in adults with ASD and SZ, (2) spatially dynamic visual-tactile facilitation impacting perception of the body boundary is affected in ASD but not SZ, and (3) body boundary perception is related to social-emotional function, but not in a way that maps on to clinical status.

show abstract

Section: Analyses and Statistical Modelingmentioning

confidence: 99%

Visual-Tactile Spatial Multisensory Interaction in Adults With Autism and Schizophrenia

et al. 2020

View full text Add to dashboard Cite

show abstract

“…, a J }. Liu et al (2017) and Tian et al (2020) have shown that CPMs can be fit to and work well for continuous and mixed types of responses. CPMs have also been shown to be consistent and asymptotically normal, with variance consistently estimated with the inverse of the information matrix under mild conditions including boundedness of the outcome variable (Li et al, 2022).…”

Section: Cumulative Probability Modelsmentioning

confidence: 99%

Addressing Detection Limits with Semiparametric Cumulative Probability Models

Tian¹,

Li²,

Shengxin³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Detection limits (DLs), where a variable is unable to be measured outside of a certain range, are common in research. Most approaches to handle DLs in the response variable implicitly make parametric assumptions on the distribution of data outside DLs. We propose a new approach to deal with DLs based on a widely used ordinal regression model, the cumulative probability model (CPM). The CPM is a type of semiparametric linear transformation model. CPMs are rank-based and can handle mixed distributions of continuous and discrete outcome variables. These features are key for analyzing data with DLs because while observations inside DLs are typically continuous, those outside DLs are censored and generally put into discrete categories. With a single lower DL, the CPM assigns values below the DL as having * The authors gratefully acknowledge CCASAnet investigators for providing data for the HIV study.

show abstract

“…< ϑ M . It should be noted that nonparametric methods of the transformation function h dominate the literature, and smooth low-dimensional spline approaches only recently received more attention; see [14] for a comparison of the two approaches.…”

Section: B Most Likely Transformationmentioning

confidence: 99%

Deep transformation models: Tackling complex regression problems with neural network based transformation models

Sick¹,

Hothorn²,

Dürr³

2020

Preprint

Self Cite

View full text Add to dashboard Cite

We present a deep transformation model for probabilistic regression. Deep learning is known for outstandingly accurate predictions on complex data but in regression tasks it is predominantly used to just predict a single number. This ignores the non-deterministic character of most tasks. Especially if crucial decisions are based on the predictions, like in medical applications, it is essential to quantify the prediction uncertainty. The presented deep learning transformation model estimates the whole conditional probability distribution, which is the most thorough way to capture uncertainty about the outcome. We combine ideas from a statistical transformation model (most likely transformation) with recent transformation models from deep learning (normalizing flows) to predict complex outcome distributions. The core of the method is a parameterized transformation function which can be trained with the usual maximum likelihood framework using gradient descent. The method can be combined with existing deep learning architectures. For small machine learning benchmark datasets, we report state of the art performance for most dataset and partly even outperform it. Our method works for complex input data, which we demonstrate by employing a CNN architecture on image data.

show abstract

An empirical comparison of two novel transformation models

Cited by 11 publications

References 21 publications

Visual-Tactile Spatial Multisensory Interaction in Adults With Autism and Schizophrenia

Visual-Tactile Spatial Multisensory Interaction in Adults With Autism and Schizophrenia

Addressing Detection Limits with Semiparametric Cumulative Probability Models

Deep transformation models: Tackling complex regression problems with neural network based transformation models

Contact Info

Product

Resources

About