Week 6 · Lab 6 · About 2 hours

Exploratory Data Analysis

Count the right unit, describe distributions, calculate correlation by hand and explain a Simpson reversal.

Start the first section Open Lab 6 instructions (PDF)

0/7 sections reviewed

Lab 6 is supplied as a PDF manual and practises the Week 5 EDA concepts with Week 6 chart choices. All data are printed in the manual and synthetic; no download is needed. These seven sections give a recomputed worked solution for every task. Work in one script, STUDENTID_Lab6.py; this site does not run Python.

Lecture-to-lab connections 7 sections

Section 1 · Task 1 · registrations, students and a safe join

Count the right thing →

The Week 5 lesson on counts and denominators becomes code: the same four registrations give a 75% registration pass rate and a 50% student pass rate.

2.4 · Tables, rows, columns and keys 2.8 · Joins: predict the result before running it 7.2 · Counts, rates and denominators

Lab 6 PDF p. 1 · Task 1 (12 minutes)

Section 2 · Task 2 · mean versus median, spread, IQR and histogram bins

Describe the distribution →

Week 2 summaries meet Week 6 histograms: one extreme value moves the mean but not the median, and bin edges change the picture but not the data.

3.2 · Mean, median and mode 3.6 · Range, fences and boxplots 8.4 · Chart choice follows the question 8.6 · Histograms: bins and bin width

Lab 6 PDF p. 2 · Task 2 (16 minutes)

Section 3 · Task 3 · steps 1–3

Calculate correlation by hand →

Pearson's r from the Week 5 lecture is computed from deviations, products and squared deviations before any code runs.

7.6 · Correlation captures a linear pattern

Lab 6 PDF p. 3 · Task 3 (first part)

Section 4 · Task 3 · steps 4–6

Correlation patterns and honest claims →

Scatter plots show what a single r can hide: a perfect U-shape has zero linear correlation, and a plotted subset is not the population behind a reported statistic.

7.5 · Match plots to variables 7.6 · Correlation captures a linear pattern 7.9 · Why stratification does not prove causation 8.4 · Chart choice follows the question

Lab 6 PDF p. 4 · Task 3 (continued)

Section 5 · Task 4 · denominators, bounds and small groups

Keep unknown outcomes separate →

The Week 5 unknown-outcome bounds and the Week 6 rule to name the denominator: one cohort gives 75% among known outcomes but 60%–80% for everyone.

7.3 · Percentage points versus relative percentage 7.4 · Unknown outcomes produce an honest range 8.5 · Counts, rates and honest wording

Lab 6 PDF p. 5 · Task 4 (10 minutes)

Section 6 · Tasks 5–6 · Simpson's paradox in course data

Compare overall and within groups →

The Week 5 course registrations, drawn as a grouped chart in Week 6, show full time leading in both course groups while part time leads overall.

7.7 · Simpson's paradox: calculate all three views 7.8 · The course-registration example 7.9 · Why stratification does not prove causation 8.12 · Pooled rates in a grouped chart

Lab 6 PDF p. 6 · Tasks 5–6 (24 minutes)

Section 7 · Task 7 · 16 students, support and readiness

Combined student analysis →

Everything together: data checks, pooled and stratified rates, weights, a common mix, one histogram and a written conclusion that separates association from causation.

7.7 · Simpson's paradox: calculate all three views 7.9 · Why stratification does not prove causation 8.6 · Histograms: bins and bin width 8.12 · Pooled rates in a grouped chart

Lab 6 PDF pp. 7–8 · Task 7 (30 minutes)

Before submitting

Explain your results to the TA during the lab: data checks, fractions with denominators, histogram counts, the hand correlation, the unknown-outcome bounds, Simpson weights and common-mix rates, your individual histogram and your 4–6 sentence conclusion. Save the completed script and calculations, and follow current course submission instructions.

Source: data-science-fundamentals-week-6-lab-manual.pdf. Original PDF instructions; worked explanations are study material.