________________ variables with extreme distributions can diminish the predictive power of regression models.

Input variables with extreme distributions can significantly affect the performance of regression models by introducing bias and affecting the coefficients estimated during model training. When input variables have extreme values, often referred to as outliers, they can disproportionately influence the regression line, leading to a model that does not accurately represent the relationship between variables for the majority of the dataset.

These extreme distributions can skew the results, making it harder for the model to generalize and predict outcomes effectively. In regression analysis, it is essential to have input variables that are relatively normally distributed or at least balanced in terms of their range to maintain model stability and accuracy.

Simple, output, and latent variables do not directly relate to this issue in the same way that input variables do. While outputs are the targets the model is predicting, and latent variables refer to not directly observed variables that might influence the results, the primary concern regarding extreme distributions impacting regression models lies mainly with the input variables utilized for model training.