You can do the same using the rdd.aggregateByKey() method.

By KEY, simultaneously calculate the SUM (the numerator for the average that we want to compute), and COUNT (the denominator for the average that we want to compute):

>>> aTuple = (0,0) # As of Python3, you can't pass a literal sequence to a function.

>>> rdd1 = rdd1.aggregateByKey(aTuple, lambda a,b: (a[0] + b, a[1] + 1),

lambda a,b: (a[0] + b[0], a[1] + b[1]))

Where the following is true about the meaning of each a and b pair above (so you can visualize what's happening):

First lambda expression for Within-Partition Reduction Step::

a: is a TUPLE that holds: (runningSum, runningCount).

b: is a SCALAR that holds the next Value

Second lambda expression for Cross-Partition Reduction Step::

a: is a TUPLE that holds: (runningSum, runningCount).

b: is a TUPLE that holds: (nextPartitionsSum, nextPartitionsCount).

Finally, in order to calculate the average for each KEY follow the steps below:

>>> finalResult = rdd1.mapValues(lambda v: v[0]/v[1]).collect()

>>> print(finalResult)

[(u'2013-09-09', 11.235365503035176),

(u'2013-09-01', 23.39500642456595),

(u'2013-09-03', 13.53240060820617),

(u'2013-09-05', 13.141148418977687),

... snip ...

]