random_samples are not uniform

[bdpedigo]

Trying to use the random_sample feature from client.materialize.query_table, but the result is definitely not a uniform random sample:

from caveclient import CAVEclient

client = CAVEclient("minnie65_phase3_v1")

sample_table = client.materialize.query_table(
    "synapse_target_predictions_ssa", random_sample=1000, desired_resolution=[1, 1, 1]
)
sample_table["post_pt_root_id"].value_counts()

gives

post_pt_root_id
864691135349000663    136
864691135404446702    136
864691135748978729    136
864691135992231361    136
864691136100920437    136
864691136390915199    136
864691136619155853    136
Name: count, dtype: int64

[bdpedigo]

@fcollman suggested merge_reference=False to see if this had to do with joins, that at least did not fix

from caveclient import CAVEclient

client = CAVEclient("minnie65_phase3_v1")

sample_table = client.materialize.query_table(
    "synapse_target_predictions_ssa",
    random_sample=1000,
    desired_resolution=[1, 1, 1],
    merge_reference=False,
).set_index("target_id")

full_table = client.materialize.query_table(
    "synapses_pni_2", filter_in_dict={"id": sample_table.index}
).set_index("id")

sample_table = sample_table.join(full_table, rsuffix="_right")
sample_table["post_pt_root_id"].value_counts()

gives

post_pt_root_id
864691135335308777    136
864691135355257807    136
864691135404851182    136
864691135463896894    136
864691135539391090    136
Name: count, dtype: int64