# The latent logarithm

**Abstract** Count or non-negative data are often log transformed to improve heteroscedasticity and scaling. To avoid undefined values where the data are zeros, a small pseudocount (e.g. 1) is added across the dataset prior to applying the transformation. This pseudocount considers neither the measured object's a priori abundance nor the confidence with which the measurement was made, making this practice convenient but statistically unfounded. I introduce here the latent logarithm, or lag. lag assumes that each observed measurement is a noisy realization of an unmeasured latent abundance. By taking the logarithm of this learned latent abundance, which reflects both sampling confidence/depth and the object's a priori abundance, lag provides a probabilistically coherent, stable, and intuitive alternative to the questionable, but conventional "log(x + pseudocount)." 

This repository is organized into three folders:
* data - Contains the data analyzed in the paper.
* paper - Contains raw LaTeX files for the paper. You are probably not interested in this.
* src - Contains the source code for lag, applying it to the data used in the paper, and for reproducing figures.

The src folder contains its own README for running the code.

Send questions/comments to surojitbiswas@g.harvard.edu