Multivariate machine learning methods are increasingly used to analyze neuroimaging data, often replacing more traditional “mass univariate” techniques that fit data one voxel at a time. In the ...functional magnetic resonance imaging (fMRI) literature, this has led to broad application of “off-the-shelf” classification and regression methods. These generic approaches allow investigators to use ready-made algorithms to accurately decode perceptual, cognitive, or behavioral states from distributed patterns of neural activity. However, when applied to correlated whole-brain fMRI data these methods suffer from coefficient instability, are sensitive to outliers, and yield dense solutions that are hard to interpret without arbitrary thresholding. Here, we develop variants of the Graph-constrained Elastic-Net (GraphNet), a fast, whole-brain regression and classification method developed for spatially and temporally correlated data that automatically yields interpretable coefficient maps (Grosenick et al., 2009b). GraphNet methods yield sparse but structured solutions by combining structured graph constraints (based on knowledge about coefficient smoothness or connectivity) with a global sparsity-inducing prior that automatically selects important variables. Because GraphNet methods can efficiently fit regression or classification models to whole-brain, multiple time-point data sets and enhance classification accuracy relative to volume-of-interest (VOI) approaches, they eliminate the need for inherently biased VOI analyses and allow whole-brain fitting without the multiple comparison problems that plague mass univariate and roaming VOI (“searchlight”) methods. As fMRI data are unlikely to be normally distributed, we (1) extend GraphNet to include robust loss functions that confer insensitivity to outliers, (2) equip them with “adaptive” penalties that asymptotically guarantee correct variable selection, and (3) develop a novel sparse structured Support Vector GraphNet classifier (SVGN). When applied to previously published data (Knutson et al., 2007), these efficient whole-brain methods significantly improved classification accuracy over previously reported VOI-based analyses on the same data (Grosenick et al., 2008; Knutson et al., 2007) while discovering task-related regions not documented in the original VOI approach. Critically, GraphNet estimates fit to the Knutson et al. (2007) data generalize well to out-of-sample data collected more than three years later on the same task but with different subjects and stimuli (Karmarkar et al., submitted for publication). By enabling robust and efficient selection of important voxels from whole-brain data taken over multiple time points (>100,000 “features”), these methods enable data-driven selection of brain areas that accurately predict single-trial behavior within and across individuals.
► We introduce robust, interpretable models for prediction with whole-brain fMRI data. ► These use a sparsity-inducing penalty that automatically selects important voxels. ► They also include a graph penalty to structure the solution. ► They outperform state-of-the-art classifiers on whole-brain fMRI data. ► They predict outcomes on new data collected years after the data used for training.
We develop a general approach to valid inference after model selection. At the core of our framework is a result that characterizes the distribution of a post-selection estimator conditioned on the ...selection event. We specialize the approach to model selection by the lasso to form valid confidence intervals for the selected coefficients and test whether all relevant variables have been included in the model.
Statistical learning and selective inference Taylor, Jonathan; Robert J. Tibshirani
Proceedings of the National Academy of Sciences - PNAS,
06/2015, Letnik:
112, Številka:
25
Journal Article
Recenzirano
Odprti dostop
We describe the problem of âselective inference.â This addresses the following challenge: Having mined a set of data to find potential associations, how do we properly assess the strength of ...these associations? The fact that we have âcherry-pickedââsearched for the strongest associationsâmeans that we must set a higher bar for declaring significant the associations that we see. This challenge becomes more important in the era of big data and complex statistical modeling. The cherry tree (dataset) can be very large and the tools for cherry picking (statistical learning methods) are now very sophisticated. We describe some recent new developments in selective inference and illustrate their use in forward stepwise regression, the lasso, and principal components analysis.
Significance Most statistical analyses involve some kind of âselectionââsearching through the data for the strongest associations. Measuring the strength of the resulting associations is a challenging task, because one must account for the effects of the selection. There are some new tools in selective inference for this task, and we illustrate their use in forward stepwise regression, the lasso, and principal components analysis.
This book provides a fresh look at the question of learner motivation and engagement, beginning with an investigation of potential motivations not to learn, the better to help instructors find more ...successful ways to engage learners in any given situation. After examining various kinds of resistance to learning, the book goes on to describe effective ways of overcoming resistance and engaging learners.Grounded in the literature of many fields, such as Adult Education, Psychology, Sociology, Cultural Anthropology, and Communication (as well as the author's own decades of experience), the book connects the concepts surrounding learning resistance directly to engagement and human motivation, drawing these ideas together to make the case for practicing motivational immediacy in all learning spaces. The second section of the book focuses on the various tools effective teachers might use to mitigate learner resistance and foster authentic and lasting engagement. The author devotes a chapter to using curriculum and Instructional Systems Design (ISD) processes to effectively foster engaged learning in different learning spaces and contexts. Two chapters are devoted to applying the theory and methods to specific domains: online learning environments, and face-to-face classrooms with both undergraduate and graduate students. The last section includes a chapter that provides a potential method to measure effectual learning in the classroom, and one that addresses the ethical issues sometimes said to exist in efforts to mitigate learner resistance and foster engagement in its place. The final chapter draws the book to a close by presenting a fluid whole that will greatly improve understanding of the ideas as well as the methods best used to reduce learning resistance, increase learner engagement, and facilitate motivational immediacy and effectual learning.
We propose new inference tools for forward stepwise regression, least angle regression, and the lasso. Assuming a Gaussian model for the observation vector y, we first describe a general scheme to ...perform valid inference after any selection event that can be characterized as y falling into a polyhedral set. This framework allows us to derive conditional (post-selection) hypothesis tests at any step of forward stepwise or least angle regression, or any step along the lasso regularization path, because, as it turns out, selection events for these procedures can be expressed as polyhedral constraints on y. The p-values associated with these tests are exactly uniform under the null distribution, in finite samples, yielding exact Type I error control. The tests can also be inverted to produce confidence intervals for appropriate underlying regression parameters. The R package
selectiveInference
, freely available on the CRAN repository, implements the new inference tools described in this article. Supplementary materials for this article are available online.
Curli are extracellular functional amyloids that are assembled by enteric bacteria during biofilm formation and host colonization. An efficient secretion system and chaperone network ensures that the ...major curli fiber subunit, CsgA, does not form intracellular amyloid aggregates. We discovered that the periplasmic protein CsgC was a highly effective inhibitor of CsgA amyloid formation. In the absence of CsgC, CsgA formed toxic intracellular aggregates. In vitro, CsgC inhibited CsgA amyloid formation at substoichiometric concentrations and maintained CsgA in a non-β-sheet-rich conformation. Interestingly, CsgC inhibited amyloid assembly of human α-synuclein, but not Aβ42, in vitro. We identified a common D-Q-Φ-X0,1-G-K-N-ζ-E motif in CsgC client proteins that is not found in Aβ42. CsgC is therefore both an efficient and selective amyloid inhibitor. Dedicated functional amyloid inhibitors may be a key feature that distinguishes functional amyloids from disease-associated amyloids.
Display omitted
•CsgC is an E. coli periplasmic protein that inhibits amyloid formation•CsgC inhibits amyloid formation at substoichiometric concentrations•CsgC prevents amyloid formation by inhibiting β-sheet transition•CsgC acts on client proteins that share a D-Q-Φ-X0,1-G-K-N-ζ-E motif
Functional amyloid formation in bacteria can assist in survival and infection, yet amyloids are also associated with many human diseases including Parkinson’s. Evans et al. report the discovery of a bacterial protein that efficiently prevents bacterial and Parkinson’s disease-associated amyloid assembly.
Optical tweezers are a highly versatile tool for exploration of the mesoscopic world, permitting non-contact manipulation of nanoscale objects. However, direct illumination with intense lasers ...restricts their use with live biological specimens, and limits the types of materials that can be trapped. Here we demonstrate an indirect optical trapping platform which circumvents these limitations by using hydrodynamic forces to exert nanoscale-precision control over aqueous particles, without directly illuminating them. Our concept is based on optically actuated micro-robotics: closed-loop control enables highly localised flow-fields to be sculpted by precisely piloting the motion of optically-trapped micro-rotors. We demonstrate 2D trapping of absorbing particles which cannot be directly optically trapped, stabilise the position and orientation of yeast cells, and demonstrate independent control over multiple objects simultaneously. Our work expands the capabilities of optical tweezers platforms, and represents a new paradigm for manipulation of aqueous mesoscopic systems.
Injuries are common in adult recreational athletes. Exercise-based injury prevention programmes offer the potential to reduce the risk of injury and have been a popular research topic. Yet, syntheses ...and meta-analyses on the effects of exercise-based injury prevention programmes for adult recreational athletes are lacking.
We aimed to synthesise and quantify the pooled intervention effects of exercise-based injury prevention programmes delivered to adults who participate in recreation sports.
Studies were eligible for inclusion if they included adult recreational athletes (aged > 16 years), an exercise-based intervention and used a randomised controlled trial design. Exclusion criteria were studies without a control group, studies using a non-randomised design and studies including participants who were undertaking activity mandatory for their occupation. Eleven literature databases were searched from earliest record, up to 9 June, 2022. The Physiotherapy Evidence Database (PEDro) scale was used to assess the risk of bias in all included studies. Reported risk statistics were synthesised in a random-effects meta-analysis to quantify pooled treatment effects and associated 95% confidence intervals and prediction intervals.
Sixteen studies met the criteria. Risk statistics were reported as risk ratios RRs (n = 12) or hazard ratios HRs (n = 4). Pooled estimates of RRs and HRs were 0.94 (95% confidence interval 0.80-1.09) and 0.65 (95% confidence interval 0.39-1.08), respectively. Prediction intervals were 0.80-1.09 and 0.16-2.70 for RR and HR, respectively. Heterogeneity was very low for RR studies, but high for HR studies (tau = 0.29, I
= 81%). There was evidence of small study effects for RR studies, evidenced by funnel plot asymmetry and Egger's test for small study bias: - 0.99 (CI - 2.08 to 0.10, p = 0.07).
Pooled point estimates were suggestive of a reduced risk of injury in intervention groups. Nevertheless, these risk estimates were insufficiently precise, too heterogeneous and potentially compromised by small study effects to arrive at any robust conclusion. More large-scale studies are required to clarify whether exercise-based injury prevention programmes are effective in adult recreational athletes.
The protocol for this review was prospectively registered in the PROSPERO database (CRD42021232697).