查找价值中位数

Question

I have a data set which is a data set of gene nodes. It has a pair of nodes and their is some value as weight given. I have to find a median of the corresponding gene pair. I count the number of times the node pair occurs in entire data set and then calculate the median of the value. Here Col[0] and Col[1] are the node pairs and Col[2] is the weight. The code below prints the nodes and the median value for odd occurrences correct but for even occurrences it shows the larger value of the two middle values. Any suggestions appreciated.

Input Type:Small list from large file.

5372 937 65.0
4821 937 65.0
4376 937 65.0
2684 937 65.0
4391 3715 1880.0
3436 1174 2383.0
3436 3031 2383.0
3436 1349 2383.0
5372 937 70.0
4821 937 70.0
4376 937 70.0
2684 937 70.0
3826 896 10.0
3826 896 17.0
5372 937 62.0
4821 937 62.0
4376 937 62.0
2684 937 62.0
3826 896 50.0
4944 3715 482.0
4944 4391 482.0
2539 1431 323.0
5372 937 59.0
4821 937 59.0
4376 937 59.0
2684 937 59.0
896 606 11.0
3826 896 10.0
5045 4901 11.0
4921 4901 11.0
4901 3545 11.0
4901 3140 11.0
4901 4243 11.0

code:

from collections import defaultdict
import numpy as np

pt  = defaultdict(float)
pm  = defaultdict(float)
pc  = defaultdict(int)
with open('input.txt', 'r') as f:
    with open('output.txt', 'w') as o:
    for numline, line in enumerate((line.split() for line in f), start=1):
        pair = line[0], line[1]
        pc[pair] += 1       
        pt[pair] = float(line[2])
        pm[pair] = np.median(pt[pair])
        print pair, pc[pair], pm[pair]

As per the definition median for even set of numbers is the average value of the two middle numbers and for odd set of number the middle value is the median. How can I get a better median value in case of even set of numbers?

Answer 1

Your pt dictionary is not right. You are storing the last weight of each pair, and to compute the median you need the whole list of weights. You could do:

from collections import defaultdict
import numpy as np

pt  = defaultdict(list)
pc  = defaultdict(int)
with open('input.txt', 'r') as f:
    with open('output.txt', 'w') as o:
    for numline, line in enumerate((line.split() for line in f), start=1):
        pair = line[0], line[1]
        pc[pair] += 1       
        pt[pair].append(float(line[2]))

# now with the medians
pm  = dict()
for pair, weights in pt.items():
    pm[pair] = np.median(weights)
    print pair, pc[pair], pm[pair]