Here we show the bias in the predicted survival probabilities when cross-entropy based loss is used regardless of the underlying network architecture. This capsule contains Python scripts for data preparation, training, and testing survival networks (e.g. DRSA and DeepHit) with cross-entropy and negative log-likelihood losses. Also, R scripts for visualizing calibration plots from the network's predictions for the CRASH2 dataset are provided.
Data and code to reproduce statistical analyses and create figures for the manuscript "Excess nutrients exacerbate heat stress to induce mass coral bleaching". The codes analyze the oceanography of the Red Sea, particularly the temporal and spatial dynamics of heat stress that causes mass coral bleaching. Additionally, other factors are assessed as potential factors in contributing to bleaching such as nutrients, photosynthetically active radiation, and the rate of warming. Finally, validations are conducted between satellite-derived sea surface temperature and in situ temperature loggers. "stress bands" that were observed within the skeletons of long-lived Porites corals collected from the Great Barrier Reef and Coral Sea. Tests performed within the code include analysis of stress band occurrence vs. time, stress band occurrence vs. heat stress during the 21st century, changes in sensitivity (stress bands / heat stress index) over time, and odds ratios of stress bands within individual cores. Additionally, the code compares satellite-derived sea surface temperature data to in situ temperature loggers for select reef locations.
Cancer drug development has been riddled with high attrition rates, in part, due to poor reproducibility of preclinical models for drug discovery. Poor experimental design and lack of scientific transparency may cause experimental biases that in turn affect data quality, robustness and reproducibility. Here, we pinpoint sources of experimental variability in conventional 2D cell-based cancer drug screens to determine the effect of confounders on cell viability for MCF7 and HCC38 breast cancer cell lines treated with platinum agents (cisplatin and carboplatin) and a proteasome inhibitor (bortezomib). Variance component analysis demonstrated that variations in cell viability were primarily associated with the choice of pharmaceutical drug and cell line, and less likely to be due to the type of growth medium or assay incubation time. Furthermore, careful consideration should be given to different methods of storing diluted pharmaceutical drugs and use of DMSO controls due to the potential risk of evaporation and the subsequent effect on dose-response curves. Optimization of experimental parameters not only improved data quality substantially but also resulted in reproducible results for bortezomib- and cisplatin-treated HCC38, MCF7, MCF-10A, and MDA-MB-436 cells. Taken together, these findings indicate that replicability (the same analyst re-performs the same experiment multiple times) and reproducibility (different analysts perform the same experiment using different experimental conditions) for cell-based drug screens can be improved by identifying potential confounders and subsequent optimization of experimental parameters for each cell line.
This material is the code associated with the paper "Adaptive Multirobot Formation to Enclose and Track a Target with Motion and Visibility Constraints" that implements the different simulation examples of the proposed multi-robot formation strategies for enclosing a moving target while maintaining motion and field-of-view constraints. Some of the parameters that can be easily modified are the selection of the target trajectory and the type of strategy to comply with motion and FOV constraints. The number of robots and the formation scale as well as the angular position of each robot in the formation can also be configured. Particular explanations are provided at the beginning of the code. Abstract of the paper: Addressing the problem of enclosing and tracking a target requires multiple agents with adequate motion strategies. We consider a team of unicycle robots with a standard camera on board. The robots must maintain the desired enclosing formation while dealing with their nonholonomic motion constraints. The reference formation trajectories must also guarantee permanent visibility of the target by overcoming the limited field of view of the cameras. We present a novel approach to characterize the conditions on the robots' trajectories taking into account the motion and visual constraints. We also propose online and offline motion planning strategies to address the constraints involved in the task of enclosing and tracking the target. These strategies are based on maintaining the formation shape with variable size or, alternatively, on maintaining the size of the formation with flexible shape.
This CodeOcean capsule is an extension of the Proof of Concept version of code to transform the online chemical disclosure site for hydraulic fracturing, FracFocus.org, into a usable database. The code performs cleaning, filtering, and curating techniques to yield organized data sets and sample analyses from a otherwise difficult to use collection of chemical records. For a majority of the records, the mass of the chemicals is calculated. The *full* data set includes all data records from the FracFocus bulk download plus flags indicating various conditions such as duplication and out-of-range values. Other generate fields are included that are our best guess fixes for the problems we detect. See the DataDictionary for a description of flags. The *filtered* data set removes problem records. To be included in *filtered* data sets, - Fracking events must use water as carrier and percentages must be consistent and within tolerance. - Chemicals must be identified by a match with an authoritative CAS number or be labeled proprietary. Further, portions of the raw bulk data that are filtered out include: - fracking events with no chemical records (mostly 2011-May 2013; but in this version, are replaced with the SkyTruth archive). - fracking events with multiple entries (and no indication which entries are correct). - chemical records that are identified as redundant within the event. Finally, I clean up some of the labeling fields by consolidating multiple versions of a single category into an easily searchable name. For instance, I collapse the 80+ versions of the supplier 'Halliburton' to a single name. By removing or cleaning the difficult data from this unique data source, I produce a data set that should facilitate more in-depth analyses of chemical use in the fracking industry. See the README file for lists of what has changed in different versions and for the data of the bulk download from FracFocus.
Scientific claims in biomedical research are typically derived from statistical analyses. However, misuse and misunderstanding of statistical procedures and results permeates the biomedical literature, affecting the validity of those claims. One approach journals have taken to address this issue is to enlist expert statistical reviewers. How many journals do this, how statistical review is incorporated, and how its value is perceived by editors is of interest. Here we report an expanded version of a survey conducted more than 20 years ago by Goodman and colleagues (1998) with the intention of characterizing contemporary statistical review policies at leading biomedical journals. We received eligible responses from 107 of 364 (28%) journals surveyed, across 57 fields, mostly from editors in chief. 34% (36/107) rarely or never use specialized statistical review, 34% (36/107) used it for 10-50% of their articles and 23% used it for all articles. These numbers have changed little since 1998 in spite of dramatically increased concern about research validity. The vast majority of editors regarded statistical review as having substantial incremental value beyond regular peer review and expressed comparatively little concern about the potential increase in reviewing time, cost, and difficulty identifying suitable statistical reviewers. Improved statistical education of researchers and different ways of employing statistical expertise are needed. Several proposals are discussed.