Application of robust statistical methods to background tracer data characterized by outliers and left-censored data

Accurate analysis of tracer-breakthrough curves is dependent on the removal of measured background concentrations from the measured tracer recovery data. Background concentrations are commonly converted to a single mean background concentration that is subtracted from tracer recovery data. To obtain an improved estimate for the mean background concentration, a statically-robust procedure addressing left-censored data and possible outliers in background concentration data is presented. A maximum likelihood estimate and other robust methods coupled with outlier removal are applied. Application of statically-robust procedures to background concentrations results not only in better estimates for mean background concentration but also results in more accurate quantitative analyses of tracer-breakthrough curves when the mean background concentration is subtracted.

► This study examines the importance of background tracer concentrations on breakthrough curves. ► Calculations consider outliers removal and outliers accommodation. ► Measured concentrations below detection limits are considered in calculations for the sample mean, median, etc. ► Various substitution methods were used in comparison with a maximum likelihood estimation. ► Results suggest the maximum likelihood estimation method may be best after outliers removal.