Correct measurement of each day an infection incidence is essential to epidemic response. Nevertheless, delays in symptom onset, testing, and reporting obscure the dynamics of transmission, necessitating strategies to take away the results of stochastic delays from noticed knowledge. Present estimators might be delicate to mannequin misspecification and censored observations; many analysts have as an alternative used strategies that exhibit robust bias. We develop an estimator with a regularization scheme to deal with stochastic delays, which we time period the strong incidence deconvolution estimator. We examine the tactic to current estimators in a simulation research, measuring accuracy in a wide range of experimental circumstances. We then use the tactic to review COVID-19 data in the USA, highlighting its stability within the face of misspecification and proper censoring. To implement the strong incidence deconvolution estimator, we launch incidental, a ready-to-use R implementation of our estimator that may assist ongoing efforts to watch the COVID-19 pandemic.