Everything so far converges here: pick the right method from your research question, know the assumptions you are signing up to, and say the sentence that wins the viva.
Almost every method choice reduces to two questions: Is there a target? and if so, what kind?
No target→ you want hidden groups → clustering · or just a pair's link → correlation.
Numbercontinuous target → linear regression.
Yes / Nobinary target → logistic regression (or a tree if prediction matters more than interpretation).
Test yourself. Read each research question, then tap the method you'd defend.
Interactive 1 — pick the method for each question. Instant feedback.
Answer the questions above…
2 · THE ASSUMPTIONS YOU'RE SIGNING
DUE DILIGENCE
Every method has a contract
Choosing a method commits you to its assumptions. Examiners probe exactly here. Pick a method to see its checklist — and the one-line answer when asked "did you check?"
Interactive 2 — pick a method; read its contract.
Pick a method…
3 · THE SENTENCE THAT WINS THE VIVA
YOUR DEFENCE IN ONE BREATH
Build your methodology sentence
A strong defence compresses into one fluent sentence: goal → outcome type → method → what you report → the limitation you own. Make the four choices and assemble yours.
Interactive 3 — choose; your sentence writes itself.
My goal is to…
…a ______ outcome.
So I use…
…reporting…
…and I acknowledge that…
Make a selection in each row…
RECAP · THE WHOLE SERIES
SIX PARTS, ONE THREAD
What you now hold
Part 1 — a model is a line; fitting means minimising squared error (least squares, R²).
Part 2 — supervised (a target) vs unsupervised (find structure); learning = lowering a loss; k-means.
Part 3 — the five methods and the equation under each.
Part 4 — reading coefficients, p-values, confidence intervals, R² and odds ratios.
Part 5 — confounding, overfitting, data quality.
Part 6 — choosing a method and defending it in one sentence.
You don't need to be a statistician. You need to choose the right tool, read what it says, name what could be wrong, and defend the choice. You can now do all four.
Scroll progress saves on this device. Part of the DBA Data-Analysis series by Jan Erik Meidell.