Statistical Modeling and Estimation of Censored Pathloss Data

TL;DR

This paper addresses the bias in pathloss parameter estimation caused by measurement censoring and proposes a Tobit maximum-likelihood estimator to improve accuracy in modeling wireless signal attenuation.

Contribution

It introduces the application of Tobit maximum-likelihood estimation for censored pathloss data, providing a method for unbiased parameter estimation.

Findings

Tobit estimator yields consistent pathloss parameter estimates.

Asymptotic variance of the estimator is derived.

Simulation results demonstrate improved accuracy over ordinary least squares.

Abstract

Pathloss is typically modeled using a log-distance power law with a large-scale fading term that is log-normal. However, the received signal is affected by the dynamic range and noise floor of the measurement system used to sound the channel, which can cause measurement samples to be truncated or censored. If the information about the censored samples are not included in the estimation method, as in ordinary least squares estimation, it can result in biased estimation of both the pathloss exponent and the large scale fading. This can be solved by applying a Tobit maximum-likelihood estimator, which provides consistent estimates for the pathloss parameters. This letter provides information about the Tobit maximum-likelihood estimator and its asymptotic variance under certain conditions.