Calculating conditional medians of a numpy array

I have a numpy array X of the form

[ A 1 a1 ]
[ A 2 a2 ]
[ A 3 a3 ]
[ B 1 b1 ]
[ B 2 b2 ]
[ B 3 b3 ]
[ B 4 b4 ]
[ C 1 c1 ]
[ C 2 c2 ]
[ C 3 c3 ]
[ C 4 c4 ]
[ C 5 c5 ]

where

(A, B, C) correspond to different experiment setups,
(1, 2, 3, ...) correspond to independent replications of the experiment, and
(a1, b2, etc.) correspond to the measurements made at each replication

I need to reduce this array to include summaries of what happened with each setup of the experiment, e.g. outputting the array

[ A median(a1, a2, a3) ]
[ B median(b1, b2, b3, b4) ]
[ C median(c1, c2, c3, c4, c5) ]

I would like to do this without having to indicate

i) how many different experiment setups there were, and

ii) how many replications of each experiment were performed.

I suspect this should be possible with some sort of masking, e.g. something like median(X[:,2] such that X[:,0] = a), iterating over a in some way, but I'm not sure of the syntax for doing so.

Solution

import numpy as np

experiments = np.unique(X[:,0])
medians = []
for experiment in experiments:
    l = [experiment]
    l.append(np.median(X[X[:,0]==experiment, 2]))
    medians.append(l)
medians = np.array(medians)