Step 5 · Analysis and line of argument (Criterion C)

Say what your results show, why, and what that means for your question.

Analysis is not a list of numbers

✗ Weak

“At α = 0.9, linear probing took 4.81 µs and chaining took 1.12 µs. At α = 0.5 linear probing took 0.62 µs and chaining took 0.71 µs.”

(true, but it just repeats the table)

✓ Strong

“Below α = 0.5 linear probing was slightly faster than chaining (0.62 vs 0.71 µs) even though it made more probes. This is likely because its keys sit next to each other in one array, so it benefits from CPU caching, while chaining follows pointers to separate lists. Above 0.7 clustering takes over: mean probes rose from 1.5 at α = 0.5 to 5.4 at α = 0.9, close to Knuth's prediction of ½(1 + 1/(1−α)) (1998).”

The hash-table essay is a made-up example used throughout this guide. Write about your own question, data and decisions.

Description or analysis? Self-check drill

Analysis explains why and says what it means. Sort these sentences.

A pattern for each finding

1 · ResultWhat happened, with numbers and a figure reference
2 · ExplainWhy, using CS theory
3 · CompareWith your prediction and published results
4 · LinkWhat this means for the RQ
Each finding Real Experts Compare Literature
Result → Explain → Compare → Link back to the RQ.

Read the shape of your data

Guess the growth rate Simulator

A classic piece of CS analysis: from timings alone, which complexity class is it? Overlay the curves, use the doubling ratios, then commit to a verdict.

Line of argument

A line of argument is the thread from research question → findings → conclusion. The reader should always know how the paragraph they're reading helps answer the RQ.

  • State your expectation in the introduction, then test it.
  • End each analysis section with one sentence that links back to the RQ.
  • Your conclusion answers the question as worded, including “to what extent”.
  • You don't need an extra section for this: connect the parts you already have.

Check your understanding Quick check

Charts that help the argument

  • Line charts for a variable that changes (size, load factor); bar charts for separate categories.
  • Error bars or the standard deviation, so the reader can see how reliable each point is.
  • A log scale when values grow by orders of magnitude, labelled as such.
  • Don't hide anomalies: point them out and explain them.
  • Tables don't count as words, but text explaining them does, and that is where the marks are.
Real life: the noise that won a Nobel Prize In 1964 Arno Penzias and Robert Wilson couldn't get rid of a faint hiss in their radio antenna. They even cleaned out pigeon droppings. Instead of ignoring the anomaly they reported it: it was the cosmic microwave background, the afterglow of the Big Bang (Nobel Prize, 1978). Investigate your outliers.
Why was the scatter graph so confident?
It had no error bars to hold it back.

Ready to move on?