Performs LASSO-penalized logistic regression for variable selection in binary classification problems. Ideal for diagnostic pathology studies that build classifiers (e.g., tumor type A vs B) with automatic feature selection.
Usage
lassologistic(
data,
outcome,
outcomeLevel,
explanatory,
penalty = "lasso",
alpha = 0.5,
lambda = "lambda.1se",
nfolds = 10,
random_seed = 123456,
standardize = TRUE,
suitabilityCheck = TRUE,
bootstrapValidation = FALSE,
bootstrapN = 200,
cv_plot = TRUE,
coef_plot = TRUE,
roc_plot = TRUE,
scoringSystem = FALSE,
scoringMethod = "schneeweiss",
scoringMaxPoints = 10,
scoreCutMethod = "median",
scoreCutPoints = "",
scoreLookupTable = TRUE,
showSummary = FALSE,
showExplanations = FALSE,
showMethodologyNotes = FALSE,
includeClinicalGuidance = FALSE,
showVariableImportance = FALSE,
showModelComparison = FALSE
)Arguments
- data
The data as a data frame.
- outcome
Binary outcome variable to classify. Can be factor or numeric with exactly two observed values.
- outcomeLevel
Level of outcome considered as the positive class (event). For factor outcomes, if left empty the second level is used.
- explanatory
Variables to be considered for selection. At least 2 required. Constant variables are removed automatically.
- penalty
Type of regularization penalty. LASSO (L1) produces sparse models by setting coefficients to zero. Ridge (L2) shrinks but retains all. Elastic Net combines both.
- alpha
Mixing parameter for elastic net. Only used when penalty is elasticnet.
- lambda
Method for selecting the optimal lambda from cross-validation. 1SE rule tends to select fewer variables (more conservative).
- nfolds
Number of folds for cross-validation. Automatically reduced when sample size is limited.
- random_seed
Random seed for reproducible cross-validation fold assignment.
- standardize
Whether to standardize predictor variables before fitting.
- suitabilityCheck
Check sample size, events-per-variable ratio, multicollinearity, and whether regularization is needed.
- bootstrapValidation
Perform bootstrap optimism-corrected validation (Harrell method). Reports corrected AUC, calibration slope, and Brier score.
- bootstrapN
Number of bootstrap iterations for internal validation.
- cv_plot
Show cross-validation deviance vs lambda plot.
- coef_plot
Show bar plot of selected variable coefficients.
- roc_plot
Show ROC curve with AUC and confidence interval.
- scoringSystem
Convert model coefficients to an integer-point scoring system. Multiple methods available following published standards.
- scoringMethod
Method for converting regression coefficients to integer points. Beta10: multiply by 10, round (Zhang et al. 2017). Schneeweiss: divide by smallest absolute coefficient (Mehta et al. 2016). Sullivan: reference-variable scaling (Sullivan et al. 2004, Framingham method). Compare: generate all three and show performance comparison.
- scoringMaxPoints
Points assigned to the strongest predictor in the reference-scaled Sullivan method. The Beta10 method uses a fixed x10 coefficient scaling and ignores this value; the Schneeweiss method scales by the smallest coefficient and also ignores it.
- scoreCutMethod
How continuous predictors are dichotomized when awarding points. A predictor earns its points when the patient's value exceeds this cut. 'median' (the default) splits at the sample median, which is data-derived and will differ in another cohort. 'manual' lets you enter established clinical thresholds, which is what makes a score portable.
- scoreCutPoints
Cut points on the ORIGINAL measurement scale, as 'variable=value' pairs separated by commas or semicolons, e.g. 'ki67=20, age=65'. Used when Cut Point method is 'manual'. Any continuous predictor not listed falls back to the sample median, and that fallback is reported. Binary predictors ignore this - they score when present.
- scoreLookupTable
Generate a lookup table mapping each possible total score to predicted probability of the positive class.
- showSummary
Display a natural-language summary of the main results.
- showExplanations
Display explanations of LASSO logistic regression methodology.
- showMethodologyNotes
Show technical notes about regularization and validation.
- includeClinicalGuidance
Include guidance for clinical interpretation of results.
- showVariableImportance
Display variable importance rankings across lambda values.
- showModelComparison
Compare LASSO vs standard logistic using selected variables.
Value
A results object containing:
results$todo | a html | ||||
results$notices | a html | ||||
results$suitabilityReport | a html | ||||
results$modelSummary | a table | ||||
results$coefficients | a table | ||||
results$performance | a table | ||||
results$scoringTable | a table | ||||
results$scoringPerformance | a table | ||||
results$methodComparison | a table | ||||
results$lookupTable | a table | ||||
results$validationTable | a table | ||||
results$cv_plot | an image | ||||
results$coef_plot | an image | ||||
results$roc_plot | an image | ||||
results$predictions | an output | ||||
results$summaryText | a html | ||||
results$lassoExplanation | a html | ||||
results$methodologyNotes | a html | ||||
results$clinicalGuidance | a html | ||||
results$variableImportance | a table | ||||
results$modelComparison | a table |
Tables can be converted to data frames with asDF or as.data.frame. For example:
results$modelSummary$asDF
as.data.frame(results$modelSummary)