Skip to content

Survey Sampling Design Studio

Simple random sampling from the frame: every listed unit has the same chance n/N_frame. Units missing from the frame can never be selected, so lowering a coverage slider shifts the target, not the precision.

n = 160, estimate 53.76, SE 1.236

Population map: 40 settlements (PSUs)

square area ∝ settlement size; pale band = units missing from the frame; thick outline = selected PSU. Arrow keys move the trace.

BudapestTownsVillages
Population map: 40 settlements (PSUs). Budapest districts: N = 655, N_h (frame) = 655; Towns: N = 885, N_h (frame) = 885; Villages: N = 487, N_h (frame) = 487. realised n: 160.

Allocation by settlement type

Budapest49 (31 %)Towns78 (49 %)Villages33 (21 %)0400
Allocation by settlement type: Budapest districts n_h = 49, Towns n_h = 78, Villages n_h = 33.

Inclusion probability trace

A unit in PSU 1 (Budapest districts, M = 117 in the frame)

  1. 1. frame:π_k = n/N = 160/2027

π_k = 0.0789 · w_k = 1/π_k = 12.67

Population, frame and sample by settlement type
settlement typeN (population)N_h (frame)S_hn_hπ_kdomain mean ± SE
Budapest districts65565513.6490.078960.4 ± 1.84
Towns88588516.2780.078952.0 ± 1.90
Villages48748713.1330.078948.1 ± 2.30
Overall2027202715.7160Σ = 160.053.8 ± 1.24

Formula

π_k = n / N_frameVar(ȳ) = (1 − n/N) S² / n

Coverage error is not sampling error: the 'frame bias' readout stays the same whatever n you choose.

Method: Seeded synthetic population (40 settlements, 3 types, ICC 0.15). π_k are exact for every design. Estimates: Hájek mean with Taylor linearisation; one-stage designs with the finite population correction (exact SRSWOR / stratified SRSWOR), two-stage designs with the ultimate-cluster with-replacement approximation (R svydesign without fpc). Neyman allocation uses the frame S_h of units, also for two-stage designs. · Synthetic, seeded data. Illustrative, not an empirical finding.

Validation: Sample estimates and domain SEs match R survey 4.2.1 (svymean, svyby) and PPS π match R sampling::inclusionprobabilities on the studio's own samples (tests/unit/survey-design-studio.test.ts). The two-stage SE is an approximation; the simulated SE is the reference for the design's true precision. · numerically validated