Solution

lower = df.describe().loc["mean"]-3*df.describe().loc["std"]
upper = df.describe().loc["mean"]+3*df.describe().loc["std"]
print(lower)
print(upper)

Now, if you haven’t seen truth series yet, here is a quick primer. We can index pandas dataframes by a truth index so that we only get rows that satisfy conditions. The first step is to get an array of True/False by condition. Let’s get an array where True means the point is within three standard deviations of the mean and false means it is outside.

(df > lower) & (df < upper) 

We can use this truth array to return only the rows that satisfy the conditions.

df = df[(df > lower) & (df < upper)].dropna()
df.describe() 

Let’s see if plotting looks better now.

sns.jointplot(df["Unemployment"],df["GDP Rate"],kind="hex")
plt.show()

It looks better now!

Let’s try this analysis with different income levels, if we use wbdata.get_incomelevel() we’ll see the different income level options.

wbdata.get_incomelevel()

Let’s look at HIC, MIC and LIC. It’s easy to find the countries associated this way:

levels = ["HIC","MIC","LIC"]
for level in levels:
    countries = [i["id"] for i in wbdata.get_country(incomelevel=level, display=False)]
    print(countries)
    print()

wbdata.get_country(incomelevel=level, display=False) returns all country dictionaries within the specified income level, as before we need to use i[“id”] to get the country ID.