Inspecting with pandas

pandas is the central library for tabular data in Python. Its core data structure is the DataFrame — a two-dimensional table with labelled rows and columns, backed by efficient NumPy arrays. The mental model is: a DataFrame is a dict of columns, where every column is a list of the same length.

Before you transform any data, the first move is always to look at it. pandas provides four methods that together answer "what do I have?"

What each method tells you

.head(n=5) returns the first n rows. It is always your first call — a sanity check that the data loaded the way you expected. Did the columns come through? Are the values in the right fields? Does anything look obviously wrong?

.shape is a tuple (rows, columns). Check this early. If you expected 10 000 rows and you have 10, something went wrong upstream. If you have 50 columns when you expected 5, you may have loaded the wrong file.

.dtypes lists each column and its inferred type. Common types:

• int64 — whole numbers
• float64 — decimals
• object — usually strings (or a mixed column)
• bool — True/False

If a numeric column shows as object, that is a red flag: numbers have been read as text, and arithmetic on them will fail. You will fix this in the next lesson.

.describe() gives count, mean, standard deviation, min, quartiles, and max for every numeric column. Glance at the min and max — extreme values often signal data entry errors or unit mismatches (a temperature in Celsius alongside a temperature in Fahrenheit, for example).

Where to go next

Now you can see what you have. Next: data quality — the four problems (missing values, duplicates, wrong types, outliers) that contaminate almost every real dataset.