Medium Data Scientist questions

34 questions · 10 points each for a correct first answer.

1. 1.Sample size for a smaller effectab-testing, power-analysismedium
2. 2.groupby with transformpandasmedium
3. 3.Reading a 95% confidence intervalconfidence-intervals, ab-testingmedium
4. 4.Testing 20 metrics at oncemultiple-testing, ab-testingmedium
5. 5.Latest order per customersql, window-functionsmedium
6. 6.Overall rate flips when splitcausal-thinking, descriptive-statisticsmedium
7. 7.Predicting a heavily skewed targetregression, feature-engineeringmedium
8. 8.Studying only active usersbias, causal-thinkingmedium
9. 9.Users needed to detect a 1 point liftab-testing, power-analysismedium
10. 10.What statistical power meanshypothesis-testing, power-analysismedium
11. 11.Comparing means with unequal varianceshypothesis-testingmedium
12. 12.Testing two categorical variableshypothesis-testingmedium
13. 13.Bonferroni threshold for ten testsmultiple-testingmedium
14. 14.What an AUC of 0.8 meansclassification-metricsmedium
15. 15.F1 from precision and recallclassification-metricsmedium
16. 16.Which prediction hurts log loss mostclassification-metrics, calibrationmedium
17. 17.RMSE with one large errorregression-metricsmedium
18. 18.Row count after a mergepandas, joinsmedium
19. 19.groupby with a missing keypandas, missing-datamedium
20. 20.Ranking with ties in SQLsql, window-functionsmedium
21. 21.NOT IN returns no rowssql, missing-datamedium
22. 22.Suspiciously strong churn featuredata-leakagemedium
23. 23.Validating a forecasting modeltime-series, model-evaluationmedium
24. 24.Making a trending series stationarytime-seriesmedium
25. 25.Why a forest beats one treeensembles, bias-variancemedium
26. 26.Lowering the boosting learning rategradient-boosting, hyperparameter-tuningmedium
27. 27.Regularisation that zeroes coefficientsregularisation, feature-selectionmedium
28. 28.PCA on unscaled featuresdimensionality-reduction, feature-scalingmedium
29. 29.Using a model on rare positivesclass-imbalance, decision-thresholdsmedium
30. 30.Brier score for two predictionscalibration, classification-metricsmedium
31. 31.Halving a confidence intervalconfidence-intervals, samplingmedium
32. 32.Lift of the top decilemodel-evaluation, marketing-analyticsmedium
33. 33.A lift that fades after week oneab-testingmedium
34. 34.PM wants to ship an unclear winab-testing, stakeholder-communicationmedium