oottest implements the out-of-treatment testing from Kuelpmann and Kuzmics (2020). Out-of treatment testing allows for a direct, pairwise likelihood comparison of theories, calibrated with pre-existing data.
You can install the development version of oottest from GitHub with:
# install.packages("devtools")
devtools::install_github("PhilippKuelpmann/oottest")Input data should be structured in the following way:
columns represent different treatments
rows represent actions
cells record the number of subjects who chose each action on each treatment
Prediction data should be structured in the following way:
columns represent different treatments
rows represent the predicted probability of each action
the different tables represent the different theories
cells record the probability of choosing an action on each treatment depending on the theory
Here is a basic example on how you can use the vuong_statistic using predictions from two theories:
library(oottest)
data_experiment <- c(1,2,3)
prediction_theory_1 <- c(1/3,1/3,1/3)
prediction_theory_2 <- c(1/4,1/4,1/2)
vuong_statistic(data_experiment, pred_I = prediction_theory_1, pred_J = prediction_theory_2)Here is a basic example how to compare three theories, using data from two treatments:
library(oottest)
treatment_1 <- c(1,2,3)
treatment_2 <- c(3,2,1)
data_experiment <- data.frame(treatment_1, treatment_2)
theory_1 <- matrix(c(1/3,1/3,1/3, 1/3, 1/3, 1/3), nrow = 3, ncol=2)
theory_2 <- matrix(c(1/4,1/4,1/2,1/2,1/4,1/4), nrow = 3, ncol=2)
theory_3 <- matrix(c(1/3,1/3,1/3, 1/4,1/4,1/2), nrow = 3, ncol=2)
theories <- array(c(theory_1,theory_2,theory_3), dim=c(3,2,3))
vuong_matrix(data_experiment, theories)