id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
cord-354200-51wk3h75	Miller, A. C.	Statistical deconvolution for inference of infection time series	2020-10-20		.txt	text/plain	8040	547	52	In this paper, we propose a statistically robust method to infer infection time series from delayed data, which we call the Robust Incidence Deconvolution Estimator (RIDE). The first class, which we term re-convolution estimators, estimate the infection curve by sampling from an assumed delay distribution and shifting observed case reports backward in timeeffectively, applying a convolution operation in reverse. The expected value of the observed data Y is a convolution of the infection time series X with the delay distribution θ; estimation of X involves the deconvolution of Y and θ. In general, we find that the model-based approaches more accurately infer the infection time series than the re-convolution and Richardson-Lucy estimators (as measured by mean squared error). As stated in the Methods section, model-based estimators start with a likelihood model for observed case data, conditioned on the underlying incidence curve.	./cache/cord-354200-51wk3h75.txt	./txt/cord-354200-51wk3h75.txt