Series.skew() for constant series returns inconsistent values

[yonil7]

I'm not sure if this is a bug, but I'v noticed a strange behavior of skew() and kurt() on constant series:

#skew:
pd.Series(3075.3, np.arange(300)).skew() #returns -105683727.49921818
pd.Series(3075.2, np.arange(300)).skew() #returns nan
pd.Series(3075.5, np.arange(300)).skew() #returns 0

#kurt:
pd.Series(3075.3, np.arange(300)).kurt() #returns 9305221026420888.0
pd.Series(3075.2, np.arange(300)).kurt() #returns -7459707710545691.0
pd.Series(3075.5, np.arange(300)).kurt() #returns 0

(using pandas 0.17.1 (from anacondas) on a 64bit intel windows 7 pc)

[jreback]

xref for rolling versions of kurt/skew: pydata#8270

we fixed this for var here: pydata#10679

interested in doing a similar fix for kurt/skew?

[jreback]

cc @jvkersch

[jvkersch]

I can pick this up some time next week, if there are no other takers.

[jreback]

@jvkersch that would be great!

[jreback]

closed by #12121