Fig 1.
X-axis shows bi-weekly data batches. Y-axis is the percentage of ChatGPT visits of total web visits.
Fig 2.
Summary of the feature selection procedure.
The gradient of plot A (left) represents the number of repetitions (out of 1000) in which each feature was selected in the cross-validation Lasso regression (10-fold for usage; 5-fold for visits and adoption). Plot B only displays those that were selected in more than 500 repetitions, i.e., the features selected for regression models. Direction of association is colored red (negative) or blue (positive).
Fig 3.
Odds ratio and standard estimates.
The odds ratio and incidence rate ratio are re-calculated in the complete dataset for better interpretability. The three plots show odds ratio of the regression for ChatGPT usage (left), the incidence rate ratio for the visits (middle) and the incidence rate ration for adoption (right), all at 95% confidence intervals. The labels on the Y-axis are sorted decreasingly according to the odds ratio or the incidence rate ratio (X-axis).
Table 1.
Regression models for usage, visits and adoption of ChatGPT.