`pd.read_csv()` loads a CSV file into a pandas data frame so its columns can be inspected, cleaned, and analyzed. Import pandas first, then pass a quoted filename: `import pandas as pd; mds = pd.read_csv('mds.csv')`. A filename alone works when the file is in the working directory; otherwise, pass a relative or full path. By default, the first row supplies column headers. section 4.2.1
Check the result before analysis. `mds.info()` reports row count, columns, non-null counts, and inferred data types; `mds.describe()` summarizes its numeric columns. For a small data frame, printing every row is practical. For a larger one, inspect selected rows with `head()` and `tail()`. Reading a file creates the data frame, but does not establish that its values or assumptions are suitable for a model. section 4.2.1 section 5.2.1
Limit the import when only part of a file is needed:
- Pass column names to `usecols`, as in `pd.read_csv('nfl_pbp.csv', usecols=['QUARTER', 'DOWN', 'TO_GO'])`. Selecting relevant play-by-play variables at load time reduces memory use and speeds loading compared with importing the entire large data set and then subsetting it. section 6.2
- Pass zero-based column positions when those are more convenient: `pd.read_csv('street_address_listing.csv', usecols=[0])` imports only the first column; `usecols=[1]` imports the second column of the population file. section 12.4.1 section 12.4.2
>>> import pandas as pd
>>> nfl_pbp = pd.read_csv('nfl_pbp.csv',
>>> usecols = ['QUARTER',
>>> 'DOWN',
>>> 'TO_GO',
>>> 'OFFENSIVE_TEAM_VENUE',
>>> 'SCORE_DIFFERENTIAL',
>>> 'PLAY_TYPE',
>>> 'YARDS_GAINED'])
The football import shows a practical `usecols` call: load only the variables needed to predict fourth-down conversion. section 6.2
Inspect before extracting leading digits
Loading `street_address_listing.csv` with `usecols=[0]` produces an `ADDRESS_NO` column with 194,836 non-null entries. Its reported `object` type matters: addresses can contain text such as suite information. Extract the first character, inspect the distinct results, discard non-digit initials, and only then convert the retained leading digits to integers for Benford analysis. A complete non-null count alone would not catch those non-digit values. section 12.4.1
