The documented percentile and median methods work on raw, unprocessed data:
raw_data = [1,1,1,1,2,3,4,4,5,5] # observations
a = np.array(raw_data)
np.median(a)
a = np.array(raw_data)
np.percentile(a, 50)
How to achieve the same outcome using summarized_data such as below (without expanding it first):
summarised_data = [[1,4],[2,1],[3,1],[4,2],[5,2]] # [[observation, number_of_responses], [...]]
That is, without performing the equivalent of:
data = [n for i in range(len(summarised_data)) for n in [summarised_data[i][0]] * summarised_data[i][1]]
