Misclassification Bias Sensitivity Analysis

Assesses how non-differential or differential misclassification of an exposure or outcome affects observed effect estimates (OR, RR, RD). Essential for studies relying on subjective classification methods (e.g., Ki-67 visual estimation, morphologic grading, IHC scoring).

Usage

misclassificationbias(
  data,
  outcome,
  outcomeLevel,
  exposure,
  exposureLevel,
  misclassType = "nondifferential",
  senExposure = 0.85,
  specExposure = 0.9,
  senExposureCase = 0.85,
  specExposureCase = 0.9,
  senExposureControl = 0.85,
  specExposureControl = 0.9,
  effectMeasure = "or",
  rangeAnalysis = TRUE,
  senRange = "0.70,0.75,0.80,0.85,0.90,0.95",
  specRange = "0.70,0.75,0.80,0.85,0.90,0.95",
  nSimulations = 10000,
  random_seed = 42
)

Arguments

data

The data as a data frame.

outcome

Binary outcome variable (e.g., recurrence, death, disease status).

outcomeLevel

Level of the outcome considered as the event.

exposure

Binary exposure or classification variable subject to misclassification (e.g., Ki-67 grade, morphologic classification, IHC scoring).

exposureLevel

Level of the exposure considered as positive/exposed.

misclassType

Non-differential: sensitivity/specificity of the classifier are the same regardless of the true outcome. Differential: error rates may differ between outcome groups.

senExposure

Sensitivity of the classification method (probability of correctly classifying a truly positive case). For Ki-67 visual estimation, typically 0.70-0.90.

specExposure

Specificity of the classification method (probability of correctly classifying a truly negative case).

senExposureCase

Sensitivity of the classifier among cases (outcome positive). Used for differential misclassification.

specExposureCase

Specificity of the classifier among cases. Used for differential misclassification.

senExposureControl

Sensitivity of the classifier among controls (outcome negative). Used for differential misclassification.

specExposureControl

Specificity of the classifier among controls. Used for differential misclassification.

effectMeasure

Effect measure to compute and bias-adjust.

rangeAnalysis

Vary sensitivity and specificity across a plausible range to show how the bias-adjusted estimate changes. Produces a contour plot.

senRange

Comma-separated sensitivity values for range analysis.

specRange

Comma-separated specificity values for range analysis.

nSimulations

Number of Monte Carlo simulations for probabilistic bias analysis. Higher values give more stable estimates but take longer.

random_seed

Random seed for reproducible simulation results.

Value

A results object containing:

results$todo					a html
results$observedTable					a table
results$biasAnalysis					a table
results$correctedTable					a table
results$rangeTable					a table
results$rangeplot					an image
results$interpretation					a html

Tables can be converted to data frames with asDF or as.data.frame. For example:

results$observedTable$asDF

as.data.frame(results$observedTable)

Examples

# \donttest{
misclassificationbias(data = data, outcome = "survival_status",
    exposure = "tumor_grade", outcomeLevel = "dead",
    exposureLevel = "high_grade",
    senExposure = 0.85, specExposure = 0.90)
#> Error: Argument 'data' must be a data frame
# }