Free lesson · Alternative data
From a sampled panel to a population estimate
Open interactive lessonPractice calculationsExplore labs
Start with the idea
An alternative dataset usually observes a selected slice of the market rather than the complete economy.
Symbols, units & horizon
- g: declared population group, such as region
- G: number of groups
- p_g: known population share, nonnegative and summing to one
- x̄_g: sampled group mean in currency per customer per month or another fixed unit
- x̂: weighted population-mean proxy in the same units
When and why to use this
Evaluate transaction, mobility or shipment panels while separating panel growth from economic growth.
An alternative dataset usually observes a selected slice of the market rather than the complete economy.
A card panel may miss cash spending, a satellite pass may miss cloudy locations, and shipping records may exclude private ports. State the population, observation unit, collection mechanism and coverage before calculating a feature. Rights to obtain, retain and redistribute records should be documented separately.
Weighting can align known group shares, but it cannot repair arbitrary selection within each group. Record entity mapping and dated coverage changes; a growing vendor panel can otherwise look like increasing economic activity.
From a sampled panel to a population estimate
- Define mutually exclusive groups covering the target population.
- Compute a mean within each sampled group using a consistent period and unit.
- Weight by population shares, not vendor sample counts, and add; report assumptions about within-group representativeness.
Population shares are .75 and .25; sampled monthly spending means are $100 and $200. Weighted proxy=.75×100+.25×200=$125 per customer, not the unweighted $150.
Apply it in a strategy
- Freeze inputs at the stated decision time and record their units.
- Evaluate transaction, mobility or shipment panels while separating panel growth from economic growth.
- Recompute the example, then change the material assumption and explain the difference.
Research deliverable
From a sampled panel to a population estimate: produce the worked calculation, a timestamped input record and a written decision addressing this limitation: Unknown or outcome-dependent selection inside groups remains biased after reweighting.
These are synthetic mechanics examples, not historical performance or paper replications. Module evidence and research boundaries record the 12 September 2026 review.
Research sources, review dates and limitations
Python implementation
Self-contained teaching example. Python 3.10+; dependencies and input conventions are shown in the code and notation. Run in your own Python environment.
# Python 3.10+; standard library and NumPy only.
# Synthetic teaching inputs; conventions and units are defined in the notation above.
import numpy as np
def population_proxy(shares,means):
p=np.asarray(shares,float); x=np.asarray(means,float)
if p.shape!=x.shape or p.ndim!=1 or np.any(p<0) or not np.isclose(p.sum(),1) or not np.isfinite(x).all(): raise ValueError('Matched group means and population shares required')
return float(p@x)
assert population_proxy([.75,.25],[100,200])==125
print(population_proxy([.75,.25],[100,200]))Continue learning
Alternative Data: Measurement, Text & Incremental Value — all lessons- From a sampled panel to a population estimate
- A reproducible dictionary score for text
- An event return needs a predeclared benchmark
- Availability delays and signal decay
- Noisy proxies and attenuation
- Measure improvement against a frozen baseline
- From forecast accuracy to a costed decision
- A data investment includes coverage, access and ongoing costs
Quantitative finance and development glossary · Python resources and libraries · Research sources and limitations