Computes three types of estimators based on classification results:
Raw proportions: Simple proportion of samples classified to each stock.
Cook's corrected estimator: Raw proportions corrected by the inverse of the
misclassification matrix (Phi_inv). This can result in negative estimates.
Cook's constrained estimator: An iterative adjustment of Cook's corrected
estimator to ensure proportions are non-negative and sum to 1.
Usage
compute_cook_estimators(class_predictions, PHIinv, np)
Arguments
- class_predictions
A numeric vector of predicted class (stock) labels for the mixed sample.
Values should be integers from 1 to np.
- PHIinv
The inverse of the misclassification matrix (Phi).
Dimensions should be np x np.
- np
Integer, the number of populations (stocks).
Value
A list containing three numeric vectors:
- raw
Raw proportions of classification.
- cook
Cook's corrected estimates.
- cook_constrained
Cook's constrained estimates (non-negative, sum to 1).