Midterm Assignment / Assessment — Conceptual Review Guide

This guide summarizes the concepts, reasoning skills, and data-analysis habits you need to demonstrate on the midterm. It does not contain answers—its purpose is to help you review the underlying ideas behind the tasks you will complete using the Blizzard salary dataset.

Use this guide alongside the midterm README and your course notes.

Understanding DataFrames, Variable Types, and .info()

You should be comfortable:

  • Reading output from:

    • .head()
    • .info()
    • .describe()

  • Determining:

    • How many rows and columns a dataset has.

    • What a row represents (the unit of observation).

    • Which variables are:

      • Numeric (e.g., salary, percent increase)
      • Categorical (e.g., performance rating)
      • Ordinal (e.g., performance ratings that have a meaningful order)

  • Explaining why a variable is treated as a specific type.

Key idea: object does not necessarily mean “string”—it often means “this should be a categorical variable.”

Describing and Comparing Distributions

You should know how to:

  • Interpret histograms, boxplots, and summary statistics.

  • Compare distributions based on:

    • Center (mean, median)
    • Spread (standard deviation, IQR)
    • Shape (skew, tails)
    • Outliers

  • Decide which visualization makes comparison easier:

    • Dodged histograms vs stacked histograms
    • Boxplots vs histograms vs density plots

Think about:

When is it easier to compare medians? When is it easier to compare proportions? When is it easier to compare counts?

Grouped Summaries and Aggregation

You should understand:

  • How groupby(...).agg(...) computes:

    • Means
    • Medians
    • Standard deviations

  • How to interpret these summaries to make statements such as:

    • Which group earns more on average?
    • Which group has more variation in salary?

Also: Given summary output with a missing value, you should be able to reason about what the missing value should be based on context.

Categorical Variables, Ordering, and Visualization

You should be able to:

  • Recognize when a categorical variable needs an explicit order using pd.Categorical(..., ordered=True).

  • Explain how ordering affects:

    • Plot appearance
    • Axis order
    • Interpretation

You should understand the difference between:

  • Count plots showing number of observations per category.
  • Proportion plots showing relative percentages within a group.

And know when to use each.

Filtering and Subsetting Data

You should understand:

  • How to filter data with:

    • ==, !=
    • & (and), | (or)

  • Why parentheses are required in filters.

  • How filtering affects:

    • The number of rows
    • The interpretation of a plot based on filtered data

  • How .sort_values(by=...) works and what it does not do.

Key understanding:

When you see fewer bars or rows than expected, think about filtering, missing data, and .dropna().

Interpreting Relationships in Plots

When describing relationships in scatterplots or other comparisons, you should be able to talk about:

  • Direction: positive, negative, none
  • Form: linear, curved, no pattern
  • Strength: strong, moderate, weak
  • Outliers: unusual points
  • Causality: observational data ≠ causal evidence

You should recognize when an interpretation is incomplete or overly confident.

Key idea: A description of a relationship should address all four of direction, form, strength, and outliers without implying causation.

Plot Critique and Code Quality

You should be able to:

  • Identify improvements to plotting code (labels, ordering, clarity).

  • Suggest ways to represent missing values:

    • Include a “Missing” category
    • Annotate counts
    • Provide summary statistics alongside plots

Focus on:

How can the plot better communicate the data honestly and clearly?

Git + Quarto Workflow: Render, Commit, Push

You should understand:

Render

  • Executes the Quarto file.
  • Produces HTML/PDF output with the latest code, text, and plots.

Commit

  • Saves a snapshot of your changed files to your local Git repository.

Push

  • Uploads your commits to GitHub so others can see them.

Critical insight:

If you don’t push, no one else can see your changes—even if you rendered and committed.

Self-Reflection (Bonus)

Be ready to:

  • Pick any course concept and explain it in your own words.
  • Show that you understand the idea, not just the procedure.
  • Describe why it matters or when it is used.

Examples include:

  • Why histograms and boxplots complement each other
  • What dummy coding does
  • Why missing data matters
  • How Git prevents version conflicts