click below
click below
Normal Size Small Size show me how
Computer Cartography
Kernel Density Estimation
Question | Answer |
---|---|
What is kernel density estimation (interpolation)? | method for discrete point data |
What is continuous data? | Exists everywhere on the surface of the Earth |
What is discrete data? | Exists at a specific location |
What happens when you apply kernel estimation to discrete data? | Get raster of cells with density values |
How do we go from discrete to continuous surface using kernel density estimation? | Place a 3-dimensional kernel function over each point so that the point is in the center of the kernel function |
How do you calculate the density for a grid cell? | Take the cumulative distance of all the 3-dimensional kernel functions that are above the center of cell |
What does it mean if you have a grid cell with a high density? | High concentration |
What are the six kernel functions? | normal, quartic, triangular, negative exponential, and uniform |
Which two functions should you use? | Normal or quartic |
What is bandwidth? | The distance from the center of function to outer edge |
What are two characteristics of a long bandwidth? | flat and generalized |
What are two characteristics of a short bandwidth? | tall and less generalized results |
How many bandwidth options are there? | 2 |
Which bandwidth option do you vary the bandwidth by specifying the minimum number of points falling under each kernel function? | Adaptive |
Which bandwidth option do you vary the bandwidth by specifying the bandwidth to a value and having all kernel functions drawn to the value? | Fixed |
If you specify a large number of points to fall under your kernel function, how will it look? | Long and more generalized |
If you specify a small number of points to fall under your kernel function, how will it look? | Short and more detailed |
Which bandwidth type is preferred? | Adaptive |
How should you set the minimum number of points that you want to fall under your kernel function? | Using a few percent of the entire points, like 2 or 3% |
Will changing the cell size change the results or how they are visualized? | How they are visualized |