In this method, each raster cell value is the sum of the contributions from all point samples that lie within a specified radius from the cell center. Then, it is weighted by the product of the kernel density function and the count at each sample. The "count" for each sample is acquired from the selected data column in the table.
So, while the Estimate method assumes every sample in the table represents a single event, the Weighted Estimate method allows each sample to represent any number of events. The result is a Heatmap that is biased towards samples where many events have been recorded.