I85PREP
⌘ K
PCED study materials
PCED study resources

PCED: Data Collection Fundamentals

PCED · Work through your exam, one topic at a time.

All guide sections

Data Collection Fundamentals

What this domain tests

Data collection fundamentals covers where data comes from and how it is gathered: source types, measurement scales, sampling methods and the biases they introduce, and the quality checks applied before any cleaning or analysis begins. Exam items ask you to classify a variable, name the sampling technique used, or explain why a huge sample can still mislead.

Core rules and concepts

  • Primary data is collected by you for the current question; secondary data was gathered earlier or by someone else (census files, APIs, third-party logs).
  • Structured data fits rows and columns (CSV, database tables); semi-structured has keys or tags (JSON, XML); unstructured has no fixed schema (free text, images, audio).
  • Measurement scales: nominal (labels, no order: city, colour), ordinal (ordered, unequal gaps: satisfaction 1-5, education level), interval (equal gaps, no true zero: Celsius, calendar year), ratio (equal gaps and a true zero: revenue, weight, counts).
  • Sampling: simple random, systematic (every k-th record), stratified (split the population into strata, sample within each), cluster (sample whole groups). Convenience sampling is cheap and usually biased.
  • Bias sources: survivor bias, non-response bias, self-selection, coverage bias (the sampling frame misses part of the population).
  • Quality dimensions: accuracy, completeness, consistency, timeliness, uniqueness, validity.

Worked example

A gym emails a satisfaction survey to 400 members and gets 90 replies. Population = all members; sample = the 90 responders; sampling frame = the email list. Members who already quit never see it, and aggrieved members reply more, so this is self-selection plus coverage bias - a bigger mailing does not fix it. "Visits per week" is ratio; "rate the equipment 1-5" is ordinal.

Traps

  • Treating ordinal survey codes as interval data and averaging them without comment.
  • Believing more data removes bias; bias is about who is measured, not how many.
  • Mixing up population, sample and sampling frame.
  • Calling JSON unstructured.