Skip to contents

Assume a limit of detection

Source: R/model_measurements.R
LOD_assume.Rd

Pathogen concentrations below a certain threshold may not be detectable and thus erroneously measured as 0. This option adjusts for a known limit of detection and includes zero measurements in the likelihood.

Usage

LOD_assume(
  limit = NULL,
  prob = 0.95,
  LOD_type = "exponential",
  drop_prob = 1e-10,
  modeldata = modeldata_init()
)

Arguments

limit

Limit of detection. The concentration below which the pathogen cannot be detected with sufficient probability, i.e. the measurement may be zero although the pathogen is present in the sample.

prob

What desired probability of detection does the limit refer to? Default is 95% (0.95): This means that the provided limit is the smallest concentration at which the pathogen can still be detected with over 95% probability.

LOD_type

The type of LOD model used. Currently, only "exponential" is supported. This models an exponentially decreasing probability of zero measurements / non-detection as a function of concentration. The exponential model can be derived from the statistical properties of dPCR, but should also work well for other quantification methods such as qPCR.

drop_prob

Probability for non-detection below which likelihood contributions of observed concentrations are dropped from LOD model. This avoids numerical issues of the LOD model at high concentrations (very small non-detection probabilities) that can otherwise affect sampling speed. Since these likelihood contributions will be virtually zero for almost all samples anyway, parameter estimates are practically not affected.

modeldata

A modeldata object to which the above model specifications should be added. Default is an empty model given by modeldata_init(). Can also be an already partly specified model returned by other EpiSewer modeling functions.

Value

A modeldata object containing data and specifications of the model to be fitted. Can be passed on to other EpiSewer modeling functions to add further data and model specifications.

The modeldata object also includes information about parameter initialization (init), meta data (.metainfo), and checks to be performed before model fitting (.checks).

Details

The limit of detection is modeled using a hurdle model. In effect, zero measurements provide a signal that the concentration in the respective sample was likely below the limit of detection, but we don't know what the exact concentration was.

The limit of detection is specific to the quantification approach and protocol. It is usually established from a dedicated lab experiment (serial dilution experiment). It his here assumed that this experiment did not cover a large fraction of the preprocessing noise to find an optimal configuration for the exponential model.

If used together with noise_estimate_dPCR(), EpiSewer will also model the effect of pre-PCR noise on the LOD. This means that the modeled LOD could be slightly higher than specified under limit, depending on the estimated pre-PCR noise.

See also

Visualize the assumed LOD as a function of concentration: plot_LOD()

Other LOD models: LOD_estimate_dPCR(), LOD_none()