Python - Mean of each value across keys in dict

Question

I am having trouble iterating across an entire dictionary to do simple summary statistics (an average) for each element of a value across keys.

My dictionary consists of keys and values that are lists of numbers:

test_dict={'NJ':[20,50,70,90,100],'NY':[10,3,0,99,57],'CT':[90,1000,2,3.4,5]}

I know that I can access the first value of each key, for instance, by doing the below, but I am having trouble with the obvious next step of adding another for loop to iterate across all elements in the values.

location1=[element[0] for element in test_dict.values()] 
location1_avg=sum(location1)/len(location1)

My ultimate goal is to have a dictionary with labels as keys (Location 1...i) and the average value across states for that location. So the first key-value would be Location1: 40, and so on.

I have the below attempt, but the error message is 'list index out of range' and i do not know how to iterate properly in this case.

for element in test_dict.values():
    avg=list()
    for nums in element[i]:
        avg[i]=sum(element[i][nums])/len(element[i][nums])

Adding desired output per requests

soln_dict={'Location1':40,'Location2':351,'Loction3':24,'Loction4':43.24,'Loction5':54}

Thank you for your help!

Answer 1

Not sure where your error lies but the i is a dead giveaway for "using indices where it's not useful / harmful".

Your problem has a straight input/output data stream, and is a perfect match for using dictionary comprehension, iterating on the key, values and rebuilding the dict with the mean as value:

test_dict={'NJ':[20,50,70,90,100],'NY':[10,3,0,99,57],'CT':[90,1000,2,3.4,5]}

result = {k:sum(x)/len(x) for k,x in test_dict.items()}

print(result)

gives:

{'CT': 220.08, 'NJ': 66.0, 'NY': 33.8}

EDIT: you seem to want a "transposed" version with anonymized keys, in that case, just use the zipped version of the values:

result = {"location{}".format(i):sum(v)/len(v) for i,v in enumerate(zip(*test_dict.values()),1)}

gives:

{'location3': 24.0, 'location5': 54.0, 'location1': 40.0, 'location2': 351.0, 'location4': 64.13333333333334}

Answer 2

You can do this:

test_dict={'NJ':[20,50,70,90,100],'NY':[10,3,0,99,57],'CT':[90,1000,2,3.4,5]}
avg=[sum(element) / len(element) for element in test_dict.values()]
print(avg) # => [66.0, 33.8, 220.08]

And for a dictionary:

test_dict={'NJ':[20,50,70,90,100],'NY':[10,3,0,99,57],'CT':[90,1000,2,3.4,5]}
avg={ k:sum(test_dict[k]) / len(test_dict[k]) for k in test_dict}
print(avg) # => {'NJ': 66.0, 'NY': 33.8, 'CT': 220.08}

Answer to the edited question:

If the arrays always have a length of 5, use this:

test_dict={'NJ':[20,50,70,90,100],'NY':[10,3,0,99,57],'CT':[90,1000,2,3.4,5]}
avg={}
for i in range(5):
  avg['Location'+str(i+1)] = sum(test_dict[k][i] for k in test_dict)/len(test_dict)
print(avg)

Output:

{'Location1': 40.0, 'Location2': 351.0, 'Location3': 24.0, 'Location4': 64.13333333333334, 'Location5': 54.0}

Answer 3

Just do :

#loop through the dictionary
for key,value in test_dict.items(): 

   #use reduce to calculate the avg
   print(key, reduce(lambda x, y: x + y, test_dict[key]) / len(test_dict[key]))

This will print :

NJ 66.0
NY 33.8
CT 220.08

Edit : As per change in OP requirements :

l = list(iter(test_dict.values()))                      #convert values to list
print(l)
#[[20, 50, 70, 90, 100], [10, 3, 0, 99, 57], [90, 1000, 2, 3.4, 5]]
d={}                                                                  #final ditionary
for i in range(len(l[0])): 
   row_list = [row[i] for row in l]                     #get values column-wise
   d['location'+str(i+1)] = sum(row_list)/len(row_list)               #calculate avg

print(d)
#{'location1': 40.0, 'location2': 351.0, 'location3': 24.0, 'location4': 64.13333333333334, 'location5': 54.0}

Note : the average you have put in question for loaction4 is wrong.

Answer 4

To keep it as simple as possible, I would I suggest:

from statistics import mean

test_dict={'NJ':[20,50,70,90,100],'NY':[10,3,0,99,57],'CT':[90,1000,2,3.4,5]}

# put the data in a list of lists
# (throw away the city names)
l = [seq for seq in test_dict.values()]


# put together 1st values, 2nd values, etc.
r = [mean(i) for i in zip(*l)]
print(r)

Which gives:

[40, 351, 24, 64.13333333333334, 54]

I divided to conquer: I turned this dictionary into a list of lists, and then used zip to put the "columns" together. Since zip expects arguments separated by a comma and not a list, I used the star operator (*) to do the conversion.

I am not sure were one should get the list of places from? Is it just Location_ + the index no? (If yes, why not leaving it in a list?)

For the mean function, see the statistics package (for Python > 3.4). Otherwise you can write your own:

mean = lambda l: reduce(lambda x, y: x+y, l) / len(l)

I took inspiration from Finding the average of a list. That is perhaps a a little cryptic and it might have been clearer to write a function without reduce, but a one-liner makes it much easier to copy and paste.

If you are in Python 3, import reduce from functools.