A test of an idea. I want to see if adding more depth cues would help visualizing a 3D scatterplot. I used a camera with short (virtual) focal length, and also use small balls instead of points to display the data, such that the size of a ball also provides depth information, which is not available if the data points are just, well, points of the same size regardless of distance from the camera.
Different from the previous experiment, the data was generated from a two-predictor linear regression model. The two predictors were generated from a uniform distribution and were uncorrelated in the population. The residuals were generated from a normal distribution.
Resources and features used: Blender Python (numpy used for generating the random numbers)
Comments