Python Tutorial
Pandas Correlations
corr() measures how numeric columns move together. 1 is a perfect positive link, -1 is negative, 0 is none.
corr
import pandas as pd
df = pd.DataFrame({
"duration": [30, 30, 45, 60, 60],
"pulse": [110, 117, 103, 102, 100],
"calories": [409, 479, 340, 282, 300],
})
print(df.corr(numeric_only=True))One Pair
print(df["duration"].corr(df["calories"]))Correlation is not causation. A strong number only means the columns move together in this dataset.
📘 Real-World Deep Dive
Knowing <strong>Pandas Corr (pandas)</strong> well is what turns pandas from a curiosity into a daily tool — you'll reach for it in nearly every real project.
Real-Life Scenario
An end-to-end usage of Pandas Corr that you'd actually see in a data pipeline or analytics notebook.
Real-Life Example
import pandas as pd
df = pd.DataFrame({
"height": [160, 170, 180, 190],
"weight": [55, 70, 80, 95],
"age": [25, 30, 35, 40],
})
print(df.corr(numeric_only=True))Expected Output
(see source)Common mistakes
- A
DataFrameindexing pattern likedf[df.col > 5]returns a copy — use.loc[row_mask, col]for assignment to avoidSettingWithCopyWarning. - Pandas infers
objectdtype for CSVs with mixed numeric/text columns; cast withpd.to_numeric/astype("category")for big speed/memory wins. df.iterrows()is O(n) and slow; iterate withdf.itertuples()or vectorise column-wise.- Treating Pandas Corr as a black box without reading the docs — the API has subtle defaults that bite when you scale.
🚀 Performance & Best Practices
- Enable the Arrow backend:
pd.read_csv("…", engine="pyarrow", dtype_backend="pyarrow")for faster, type-stable reads. - Use
categoricaldtype for columns with low-cardinality strings — sort/join/group-by speed up dramatically. - Switching a hot loop from row-wise Python to
df.eval("…")/df.query("…")often gives 5–50×. - When working with pandas, prefer vectorised / batched operations over Python loops.
🧪 Try It Yourself
- Reproduce the snippet on a representative slice of your own data.
- Profile the snippet with
cProfileortimeitand find the single biggest improvement. - Generalise the snippet into a small, reusable function you can drop into future projects.