Section D.1 Data
Rather than download the whole dataset, we can get just an excerpt.
download("https://github.com/AllenDowney/ThinkStats/raw/v3/data/LLCP2022.hdf5")
Here are the first few rows.
$ brfss = pd.read_hdf("LLCP2022.hdf5", key="brfss")
brfss.describe()
_SEX HTM4 WTKG3 _LLCPWT
count 445132.000000 416480.000000 403054.000000 445132.000000
mean 1.529942 170.269057 8307.447039 594.856344
std 0.499103 10.717750 2144.817270 1134.837415
min 1.000000 91.000000 2268.000000 0.020464
25% 1.000000 163.000000 6804.000000 115.885991
50% 2.000000 170.000000 8074.000000 274.632388
75% 2.000000 178.000000 9525.000000 627.913694
max 2.000000 241.000000 29257.000000 54390.520926
The dataset includes self-reported heights for almost 200,000 men and almost 220,000 women. Here are numbers of males and females with the shortest recorded heights.
$ xtab = pd.crosstab(brfss["HTM4"], brfss["_SEX"])
xtab.columns = ["Male", "Female"]
xtab.head()
Male Female
HTM4
91.0 7 17
92.0 0 1
95.0 1 0
97.0 1 3
99.0 1 0
And the tallest recorded heights.
$ xtab.tail()
Male Female
HTM4
226.0 9 2
229.0 4 1
234.0 4 0
236.0 1 0
241.0 4 1
According to the codebook, heights below 91 cm are set to 91 cm, and heights above 241 cm are set to 241 cm. Even so, some of the remaining values are likely to be errors. For example, the current tallest woman in the world is 234 cm, so the largest female height in the dataset, 241 cm, was probably reported or recorded incorrectly.
To draw a bootstrap sample from this
DataFrame, we’ll use the following function.
def resample(df):
"""Draw a bootstrap sample.
df: DataFrame
returns: DataFrame
"""
n = len(df)
return df.sample(n, replace=True, weights="_LLCPWT")
Here’s a single sample we’ll use for testing.
sample = resample(brfss)
