Conversation
dougj892
left a comment
There was a problem hiding this comment.
Hey - Just had one small comment
| return gdf["neighbor_count"].values | ||
|
|
||
|
|
||
| def randomly_sample_rooftops(rooftop_neighbour_count: list | np.ndarray, |
There was a problem hiding this comment.
Hi @poornimaramesh - It looks like this function samples rooftops with probability proportional to the number of neighbors. In our testing in India, we actually found that using simple random sampling within PSUs slightly outperforms other more complicated forms of sampling in terms of ensuring equal probability of household selection and reduction of bias. The difference in performance was very minor, but considering SRS is a bit easier for folks to understand and implement I would prefer for it to be the default. Would it be possible to either a) change the name of this function to make clear that this is implementing sampling using prob proportional to number of neighbors or b) change to SRS?
No description provided.