Predicted Alzheimer's disease probabilities and observed diagnoses from a
logistic regression model fitted to a training subset of the ad_data
dataset. The model uses age, sex, cerebrospinal fluid (CSF) tau and
phosphorylated tau concentrations, amyloid beta 42 concentration, and
Apolipoprotein E (APOE) genotype as predictors.
Format
ad_validation_logistic
A data frame containing one row for each observation in the validation subset, with 3 columns:
- .pred_impaired
Predicted probability that the subject is cognitively impaired.
- .pred_control
Predicted probability that the subject is a cognitive control.
- Class
Observed diagnosis of the subject, with levels
impairedandcontrol.
The probabilities are generated from the logistic regression model:
diagnosis ~ age + male + tau + p_tau + Ab_42 + GenotypePredictor variables include:
- age
Normalized patient age.
- male
Indicator of reported male sex.
- tau
Concentration of total tau protein in cerebrospinal fluid (CSF), measured by lumbar puncture.
- p_tau
Concentration of phosphorylated tau protein in CSF, measured by lumbar puncture.
- Ab_42
Concentration of amyloid beta 42 (Aβ42) in CSF. Aβ42 is a 42-amino-acid peptide derived from amyloid precursor protein and is a component of the amyloid plaques associated with Alzheimer's disease.
- Genotype
APOE allele pair. Possible values are
E2E2,E2E3,E3E3,E3E4,E2E4, andE4E4.
Details
The original ad_data observations were split by alternating rows, with
even-numbered observations used for model training and odd-numbered
observations used for validation. The logistic regression model was fitted
to the training observations, and predicted probabilities were generated
for the validation observations.