Top 10 Statistics Concepts for Pune Data Sci (2026)

The list

1. 1

   Mean, Median, Mode + Standard Deviation

   Mean: sum / count (sensitive to outliers). Median: middle value when sorted (robust to outliers). Mode: most frequent value. Standard deviation: average distance from mean — measures spread. Use median over mean for skewed data; use std dev to quantify variability.

   Why it matters: Asked at ~90% of Pune data fresher rounds. Foundation question; expected to know cold + apply correctly.

   Best for: Foundation; always asked.

2. 2

   Normal Distribution + 68-95-99.7 Rule

   Bell-shaped + symmetric around mean. 68% of values within 1 std dev, 95% within 2 std dev, 99.7% within 3 std dev. Many real-world phenomena approximate normality (heights, errors, sums of many small effects via CLT). Z-scores translate any normal distribution to standard normal.

   Why it matters: Asked at ~70% of Pune rounds. Z-score calculation appears in technical screens.

   Best for: Probability + outlier detection foundation.

3. 3

   Central Limit Theorem (CLT)

   Sample means of any underlying distribution (with finite variance) approach a normal distribution as sample size grows — typically n≥30 is sufficient. Why it matters: lets you use normal-based inference (t-tests, confidence intervals) even when underlying data isn't normal.

   Why it matters: Asked at ~50% of Pune product company + analytics consultancy rounds. The conceptual bridge from descriptive to inferential statistics.

   Best for: Inferential statistics foundation; senior-fresher signal.

4. 4

   Hypothesis Testing (p-value, null + alternative hypothesis)

   Frame: H0 (status quo) vs H1 (effect exists). Compute test statistic + p-value. If p < α (typically 0.05), reject H0 in favour of H1. p-value = probability of seeing this result or more extreme IF H0 is true. Crucial: p-value is NOT the probability that H0 is true.

   Why it matters: Asked at ~75% of Pune rounds. Misinterpreting p-value is the most-failed statistics question — be precise.

   Best for: A/B test design + inferential statistics; universal data interview question.

5. 5

   Type I vs Type II Errors + Power

   Type I (α): rejecting true null hypothesis (false positive). Type II (β): failing to reject false null hypothesis (false negative). Power = 1 - β = probability of correctly rejecting false null. Trade-off: reducing α increases β (and vice versa). Power increases with sample size + effect size.

   Why it matters: Asked at ~45% of Pune rounds. Walk through an A/B test design where Type II error matters (missing a real product improvement).

   Best for: Experimental design depth; product-company differentiator.

6. 6

   Linear Regression + Coefficients

   Predict continuous outcome from one or more features. Equation: y = β0 + β1*x1 + β2*x2 + ... Each coefficient measures the change in y per unit change in xi (holding others constant). Diagnostics: R² (variance explained), residual analysis, multicollinearity (VIF).

   Why it matters: Asked at ~65% of Pune data scientist + analyst rounds. The starting point of supervised learning.

   Best for: ML foundation; expected fluency at every data tier.

7. 7

   Logistic Regression + Classification Metrics

   Predict binary outcome (0/1) using logistic function applied to linear combination of features. Output is probability; threshold (typically 0.5) converts to class prediction. Classification metrics: accuracy (misleading on imbalanced data) + precision + recall + F1 + ROC-AUC.

   Why it matters: Asked at ~55% of Pune rounds. Precision/recall trade-off + when to use each is the senior-fresher discriminator.

   Best for: Classification foundation; differentiates beyond pure-Analyst tier.

8. 8

   Correlation vs Causation

   Correlation: statistical association (Pearson r for linear, Spearman for rank-based). Causation: one variable directly affects another (requires controlled experiments or strong observational design). Spurious correlations are common — don't infer causation from correlation alone.

   Why it matters: Asked at ~40% of Pune rounds. Pune interviewers test critical thinking via 'what would you conclude from X correlation' questions.

   Best for: Analytical thinking signal; differentiates senior-fresher candidates.

9. 9

   Sampling Methods + Bias

   Random sampling: equal probability for every element. Stratified: sample proportionally from sub-groups. Systematic: every nth element. Cluster: sample whole groups. Bias sources: selection bias, survivorship bias, response bias, observer bias. Modern A/B tests rely heavily on randomisation to avoid bias.

   Why it matters: Asked at ~35% of Pune rounds, especially at A/B testing-heavy roles (consumer SaaS, product companies). Bias detection is a high-leverage senior-fresher topic.

   Best for: Experimental design + survey work depth.

10. 10

    Confidence Intervals + Margin of Error

    Range that's likely to contain a population parameter with stated confidence (typically 95%). Width depends on sample size, variability, confidence level. Wider intervals = less precise estimates. Crucial: 95% CI means 'if we repeated this study many times, 95% of computed intervals would contain the true parameter' — NOT 'there's a 95% chance the true value is in this interval'.

    Why it matters: Asked at ~40% of Pune rounds. Common pairing with hypothesis testing questions.

    Best for: Inference depth; precise framing signals statistical maturity.