summary.ehr2pheno: Summarize a phenotype prediction
In borishejblum/phenotypr: Elastic-net logistic regression for phenotyping prediction using EHR data

summary.ehr2pheno

R Documentation

Summarize a phenotype prediction

Summarize a phenotype prediction from diagnosis codes and NLP processed notes

## S3 method for class 'ehr2pheno'
summary(object, goldstandard = NULL, ...)

`object`	An object of class `"ehr2pheno"`
`goldstandard`	a binary vector containing the gold-standard phenotypes
`...`	Arguments to be passed to or from other methods

Computes the True Positives Rate (TPR), False Positives Rate (FPR), and Area Under the ROC Curve (AUC).

WARNING: the criteria are computed on the training data itself and are therefore over-optimistic !

an object of class "summary.ehr2pheno" with the following elements:

criteria:
- AUC: a vector of length 3 containing the AUC on the training data for codes only, nlp only, and both codes and nlp.
- TPR: a matrix of 3 columns the TPR for each threshold on the training data for codes only, nlp only, and both codes and nlp.
- FPR: a matrix of 3 columns the FPR for each threshold on the training data for codes only, nlp only, and both codes and nlp.
pred_proba: a list with the predicted probabilities for each of the three models with codes, nlp and both
pred_proba: a list with the selected features and associated odds ratios for each of the three logistic regression models with elastic-net codes, nlp and both