Explanation:
This type of sampling is where you break the data into groups where there isn't any overlap. Then you select samples from each group. You may select the same amount from each group, or perhaps you select the items based on a weighted proportion. Note how one entire group is not sampled.
Contrast this with cluster sampling where again the data is broken into non-overlapping groups; however, this time we only select a few groups to sample every member of those groups selected. This time entire full groups are sampled.
Here's an example of each