Sampling Theory and Design of Experiments
CSIR NET Mathematical Sciences Unit 4: SRS, stratified, systematic and PPS sampling, ratio and regression estimators, CRD, RBD, LSD, BIBD and confounding.
- 5
- Topics
- 98
- Min read
- 01 Simple Random, Stratified and Systematic SamplingStart here CSIR NET sampling theory: SRSWR and SRSWOR, finite population correction, stratified allocation (proportional, Neyman, cost-optimum) and systematic sampling.
- 02 PPS Sampling, Ratio and Regression Estimators CSIR NET sampling: PPS selection, Hansen-Hurwitz and Horvitz-Thompson estimators, Yates-Grundy variance, and ratio, product and regression estimators.
- 03 CRD, RBD and Latin Square Designs CSIR NET design of experiments: principles, CRD, RBD and Latin square models, ANOVA tables, missing plot formulas, efficiency and orthogonal Latin squares.
- 04 Block Designs: Connectedness, Orthogonality and BIBD CSIR NET block designs: incidence and C-matrix, connectedness, orthogonality, BIBD parameter relations, Fisher's inequality and intrablock analysis.
- 05 2^k Factorial Experiments and Confounding CSIR NET factorial designs: 2^2 and 2^3 effects, Yates' algorithm, sums of squares, total and partial confounding, principal block and generalized interaction.
- Sampling and Design of Experiments Formula Sheet and Short Notes Revise The whole chapter on one page.
Chapter Overview
Sampling theory and design of experiments sit in Unit 4 (statistics) of the CSIR NET Mathematical Sciences syllabus. They answer two sides of one question: how should data be collected so that a simple estimate is precise?
- In sampling, the population is fixed and finite, and all randomness comes from the way units are drawn. The tools are inclusion probabilities, the finite population correction, stratification, systematic selection, selection with probability proportional to size, and estimators that borrow strength from an auxiliary variable.
- In design of experiments, we assign treatments to units ourselves. The tools are randomization, replication and local control, and the designs are the completely randomized, randomized block and Latin square designs, incomplete block designs (above all the BIBD), and \( 2^k \) factorials run in small blocks by confounding.
The two halves share one idea. Stratification in a survey and blocking in an experiment both remove variation between groups of alike units from the error, and both are analysed by splitting a sum of squares into between and within parts. Once you see that, the chapter becomes a short list of formulas and a few structural facts.
The notes assume an MSc-level first course in sampling and design (books such as Cochran’s Sampling Techniques, Sukhatme and Sukhatme’s Sampling Theory of Surveys with Applications, Montgomery’s Design and Analysis of Experiments, or Dey’s Theory of Block Designs). They state each result exactly, derive the ones Part C tests, and check every number by computation.
What You’ll Learn
By the end of this chapter you will be able to:
- Compute and compare variances under SRSWR, SRSWOR, stratified (proportional, Neyman, cost-optimum) and systematic sampling
- Select units with probability proportional to size, and use the Hansen-Hurwitz and Horvitz-Thompson estimators with their variances
- Decide when ratio, product and regression estimators beat the sample mean
- Write the ANOVA of a CRD, RBD or Latin square, estimate a missing plot and judge relative efficiency
- Use the C-matrix to decide connectedness and orthogonality of a block design, and analyse a BIBD
- Estimate \( 2^k \) factorial effects by Yates’ algorithm, and construct and analyse confounded and partially confounded arrangements
Prerequisites
- Descriptive statistics: variance, correlation, coefficient of variation, the least-squares slope (see Descriptive Statistics and Exploratory Data Analysis)
- Elementary probability on finite sample spaces (see Sample Space, Independence and Bayes’ Theorem)
- Linear algebra: rank, row space, eigenvalues of \( aI + bJ \) (see Linear Algebra)
- Arithmetic speed: Part B numericals reward the same clean fraction work as Part A Numerical Ability
Learning Path
graph TD
A[1. Simple Random, Stratified and Systematic Sampling] --> B[2. PPS Sampling, Ratio and Regression Estimators]
A --> C[3. CRD, RBD and Latin Square Designs]
C --> D[4. Block Designs: Connectedness, Orthogonality and BIBD]
C --> E[5. 2^k Factorial Experiments and Confounding]
D --> E
B --> F[Quick Revision]
E --> FPage 1 sets up the finite-population variance arithmetic that page 2 builds on. Page 3 introduces the linear model and ANOVA that pages 4 and 5 generalise: page 4 to blocks smaller than the number of treatments, page 5 to treatments with factorial structure in small blocks.
Topics in This Chapter
| Topic | What it covers | Time |
|---|---|---|
| Simple Random, Stratified and Systematic Sampling | SRSWR and SRSWOR, fpc, sample size, stratified estimator, proportional, Neyman and cost-optimum allocation, linear and circular systematic sampling | 40 min |
| PPS Sampling, Ratio and Regression Estimators | Cumulative total and Lahiri selection, Hansen-Hurwitz, Horvitz-Thompson, Yates-Grundy, ratio, product and regression estimators and their comparison | 40 min |
| CRD, RBD and Latin Square Designs | Principles of design, models and ANOVA, missing plot technique, relative efficiency, counting Latin squares, orthogonal and Graeco-Latin squares | 40 min |
| Block Designs: Connectedness, Orthogonality and BIBD | Incidence and C-matrix, estimability, connectedness, orthogonality, BIBD relations, Fisher’s inequality, intrablock analysis, existence | 40 min |
| 2^k Factorial Experiments and Confounding | Effects and contrasts, Yates’ algorithm, sums of squares, total and partial confounding, principal block, generalized interaction | 40 min |
| Quick Revision | Every formula and benchmark value on one page | 10 min |
Official syllabus lines and where they are covered
| Official topic line (Unit 4) | Covered on |
|---|---|
| Simple random, stratified, systematic sampling | Simple Random, Stratified and Systematic Sampling |
| PPS sampling; ratio and regression estimators | PPS Sampling, Ratio and Regression Estimators |
| CRD, RBD, Latin square designs | CRD, RBD and Latin Square Designs |
| Connectedness/orthogonality of block designs, BIBD | Block Designs: Connectedness, Orthogonality and BIBD |
| 2^k factorial experiments, confounding | 2^k Factorial Experiments and Confounding |
How CSIR NET Asks It
Per the NTA information bulletin for December 2026, the Mathematical Sciences paper (code 704) is marked as follows:
| Part B | Part C | |
|---|---|---|
| Questions given | 40 | 60 |
| Questions to attempt | 25 | 20 |
| Marks per correct answer | 3 | 4.75 |
| Marks per wrong answer | −0.75 | 0 |
| Maximum | 75 | 95 |
NTA publishes no topic-wise weightage, so we make no claim about how many questions come from this chapter. The formats it suits are clear:
- Part B (single correct) rewards one clean computation: an SRSWOR variance, a Neyman allocation, a sample size, a Horvitz-Thompson estimate, a missing plot value, a BIBD parameter, an effect or SS from a \( 2^k \) table. Wrong options usually come from a neighbouring convention: \( \sigma^2 \) for \( S^2 \), proportional for Neyman, \( 2p_i \) for \( \pi_i \), \( r \) and \( t \) swapped, divisor \( r2^k \) for \( r2^{k-1} \).
- Part C (one or more correct) rewards structural facts: “systematic sampling beats SRS iff…”, “the ratio estimator is better than the mean iff…”, “this design is connected but not orthogonal”, “this parameter set cannot be a BIBD”, “these effects are confounded”. Every topic page ends with Level 3 practice in exactly this format, with every option justified.
For the marking scheme and the other units, see the Mathematical Sciences overview. Found an error? See About for how to report it.
Start Learning
Begin with the first topic: