Distributional conformal prediction

We propose a robust method for constructing conditionally valid prediction intervals based on models for conditional distributions such as quantile and distribution regression. Our approach can be applied to important prediction problems, including cross-sectional prediction, k–step-ahead forecasts, synthetic controls and counterfactual prediction, and individual treatment effects prediction. Our method exploits the probability integral transform and relies on permuting estimated ranks. Unlike regression residuals, ranks are independent of the predictors, allowing us to construct conditionally valid prediction intervals under heteroskedasticity. We establish approximate conditional validity under consistent estimation and provide approximate unconditional validity under model misspecification, under overfitting, and with time series data. We also propose a simple “shape” adjustment of our baseline method that yields optimal prediction intervals.

149. Extraction-free RT-PCR to Detect SARS-CoV-2 Variants of Concern

Background SARS-CoV-2 variants of concern (VOC) have challenged real-time reverse transcriptase polymerase chain reaction (RT-PCR) methods for the diagnosis of COVID-19. Methods The CDC 2019-Novel Coronavirus real-time RT-PCR panel was modified to create a single-plex extraction-free proxy RT-PCR assay, VOCFast™. This assay uses the nucleocapsid N1 as well as novel primer/probe pairs to target VOC mutations in the Orf1a and spike (S) genes. For analytical validation of VOCFast, synthetic controls for the Wuhan, alpha/B.1.1.7, beta/B.1.351, and gamma/P.1 strains were tested at various concentrations. Clinical validation was performed using patient anterior nares swab and saliva specimens collected in the Denver, CO area between Nov 2020 and Feb 2021 or in March 2021. Orthogonal next-generation sequencing (NGS) was also performed. Results Similar N1 quantification cycle (Cq) values corresponding to viral load were observed for all strains, suggesting that VOC mutations do not affect performance of the N1 primer/probe. Orf1a-mut and S1-mut primer/probes generated a stable high Cq value for the Wuhan strain. Conversely, Orf1a-mut Cq values were inversely correlated with viral load for all VOC. The S1-mut Cq was inversely correlated with viral load of the alpha strain, but did not reliably amplify beta/gamma VOC. The limit of detection was 8 copies/uL. The first set of COVID-19 patient specimens revealed no amplification using Orf1a-mut whereas 53% of specimens collected in Mar 2021 demonstrated amplification by Orf-1a. Orthogonal testing by the SARS-CoV-2 NGS Assay and COVID-DX software demonstrated that 12/12 alpha strains, 2/2 beta/gamma strains, and 33/33 Wuhan strains were correctly identified by VOCFast. Detection of VOC in clinical specimens and validation by NGS Conclusion The combination of the N1, Orf1a-mut, and S1-mut primers/probes in VOCFast can distinguish the Wuhan, alpha, and beta/gamma strains and it consistent with NGS results. Testing of clinical samples revealed that VOC emerged in Denver, CO in March 2021. Future work to discriminate beta, gamma, and emerging VOC is ongoing. In summary, VOCFast is an extraction-free RT-PCR assay for nasal swab and saliva specimens that can identify VOC with a turnaround time suitable for clinical testing. Disclosures Brian L. Harry, MD PhD, Summit Biolabs Inc. (Grant/Research Support, Shareholder) Mara Couto-Rodriguez, MS, Biotia (Employee) Dorottya Nagy-Szakal, MD PhD, Biotia Inc (Employee, Shareholder) Niamh B. O’Hara, PhD, Biotia (Board Member, Employee, Shareholder) Shi-Long Lu, MD PhD, Summit Biolabs Inc. (Grant/Research Support, Shareholder)

Wildfires Vegetation Recovery through Satellite Remote Sensing and Functional Data Analysis

In recent years, wildfires have caused havoc across the world, which are especially aggravated in certain regions due to climate change. Remote sensing has become a powerful tool for monitoring fires, as well as for measuring their effects on vegetation over the following years. We aim to explain the dynamics of wildfires’ effects on a vegetation index (previously estimated by causal inference through synthetic controls) from pre-wildfire available information (mainly proceeding from satellites). For this purpose, we use regression models from Functional Data Analysis, where wildfire effects are considered functional responses, depending on elapsed time after each wildfire, while pre-wildfire information acts as scalar covariates. Our main findings show that vegetation recovery after wildfires is a slow process, affected by many pre-wildfire conditions, among which the richness and diversity of vegetation is one of the best predictors for the recovery.

Employment impacts of the San Francisco sugar-sweetened beverage tax 2 years after implementation

Introduction Sugar-sweetened beverage (SSB) taxes have been implemented worldwide to raise revenue and reduce consumption of SSBs, which is associated with health harms. Empirical evaluations have found that these taxes are successful at reducing demand for SSBs; however, SSB taxes face opposition, in part because of claims that they will lead to substantial job losses. The purpose of this study is to examine the impact of the San Francisco SSB tax, implemented on January 1st, 2018, on employment. Methods Monthly employment counts were obtained from the Bureau of Labor Statistics from January 2013 (5-years pre-tax) through December 2019 (2-years post-tax) for the overall economy, private sector, supermarkets and other grocery stores, convenience stores, limited-service restaurants, and beverage manufacturing. A synthetic control analysis was conducted for each employment outcome. The synthetic controls (i.e., estimated counterfactuals) were generated from a pool of urban control counties using pre-tax labor market-related characteristics. Results The synthetic controls had similar labor market-related characteristics and employment outcomes to those in San Francisco in the pre-tax period. Up to 2 years post-tax, differences in employment between San Francisco and the synthetic controls were small and not “statistically significant” based on placebo tests for all employment outcomes. Conclusions Up to two years post-tax, we do not find evidence that the San Francisco SSB tax negatively impacted net employment, employment in the private sector, or employment in specific SSB-related industries.

Using Synthetic Controls: Feasibility, Data Requirements, and Methodological Aspects

Probably because of their interpretability and transparent nature, synthetic controls have become widely applied in empirical research in economics and the social sciences. This article aims to provide practical guidance to researchers employing synthetic control methods. The article starts with an overview and an introduction to synthetic control estimation. The main sections discuss the advantages of the synthetic control framework as a research design, and describe the settings where synthetic controls provide reliable estimates and those where they may fail. The article closes with a discussion of recent extensions, related methods, and avenues for future research. (JEL B41, C32, C54, E23, F15, O47)