Statistical Analysis for Assessing Highway Maintenance Level of Service

Many transportation agencies have a maintenance quality assurance program and most use level of service (LOS) to measure the condition of highway features or the performance of maintenance services. If the necessary condition or performance data cannot be obtained from established inventory databases, most agencies use sampling techniques to collect what they need. Many agencies may be able to reduce the number of samples they collect or greatly improve the precision of the condition estimates by employing stratified sampling techniques rather than simple random sampling. The analysis of stratified samples depends on characteristics of the measure being collected. The most common types of measures for estimating LOS for maintenance quality assurance programs are simple binomial, domain binomial, and ratios in which the LOS scale is based on percentage of inventory in deficient maintenance condition. This paper, which is based on NCHRP Project 14-25, presents a statistical analysis for estimating the deficiency rate and confidence interval. Having an adequate sample size is a challenge for some measures. This paper introduces a visual tool for assessing the adequacy of the margin of error associated with LOS estimates.