Skip to main content
SLU:s publikationsdatabas (SLUpub)

Rapport2025Öppen tillgång

Analyzing temporal trends in data series variables with values below reporting limits : a case study of groundwater Pb concentrations

von Brömssen, Claudia; Olsson, Erik

Sammanfattning

Values below reporting limit1 (RL) are common when levels of the observed variable are small. In order to allow handling of such values it is often chosen to substitute values with a fixed value, typically half of the used limit. If the goal is to use the data further in statistical models, this procedure is satisfactory, as long as the number of values below a limit is small. There are, however, a few situations where it would be advantageous to retain more of the available information in the data. This is true especially if the RL changes, which is common in environmental monitoring when chemical analysis instruments or procedures are improved over time. An improved strategy for analysing such data is also valuable if a large percentage of values lie below the limit.

Different types of censored regression has been suggested to handle data under a RL. In these models, observations below the RL are not substituted with a constant, but instead the information of the limit itself and the proportion of observations below this limit are used in the calculation of the model estimates.

In this report we describe different ways to handle values below a reporting limit, which are readily available in R (R Core Team, 2024), give examples of R code and describe the procedure of analysing time series of Pb in groundwater. Values under the RL will also be called censored values.

Publicerad i

Rapport (Institutionen för energi och teknik, SLU)
2025, nummer: 129
Utgivare: Department of Energy and Technology, Swedish University of Agricultural Sciences

SLU författare

UKÄ forskningsämne

Miljövetenskap
Sannolikhetsteori och statistik

Permanent länk till denna sida (URI)

https://res.slu.se/id/publ/140363