Biostatistics · Introductory

Collection and Compilation of Data

Written by SignalsLab

Early edition: these lessons are being reviewed.

Content

Summary

This lesson covers the distinction between population variation and estimate precision via the standard error, the computation of Hedges' g for meta-analysis, and the impact of fixed versus random-effects models. It also emphasizes that rigorous study design, including blinding and proper data management, is essential for valid causal inference. These concepts collectively ensure that statistical estimates are precise and that evidence synthesis accounts for both within-study and between-study variance.

Intuition

Think of statistical inference as separating the noise in individual measurements from the reliability of your average. The standard deviation tells you how spread out individual data points are, while the standard error tells you how much your sample mean might vary if you repeated the experiment; crucially, this error shrinks as you collect more data. When combining studies in meta-analysis, we use Hedges' g to standardize differences, but we must choose between fixed-effect and random-effects models. The fixed-effect model assumes all studies share one true effect, whereas the random-effects model acknowledges that studies differ (heterogeneity), leading to wider confidence intervals and different weights. Finally, no amount of statistical tweaking can fix poor data collection, making proper study design and data integrity foundational.

Key equations

SE=sdn
(1)

Standard error of the mean, where sd is the standard deviation and n is the sample size.

Swithin=(n1−1)sd12+(n2−1)sd22n1+n2−2
(2)

Pooled within-groups standard deviation for treated (1) and control (2) groups.

d=x‾1−x‾2Swithin
(3)

Standardized mean difference (effect size), where x‾ represents the group means.

Derivation

  1. The standard error of the mean is derived by dividing the standard deviation by the square root of the sample size, showing that precision increases with larger n.
  2. Hedges' g is computed by first calculating the pooled within-groups standard deviation using the sample sizes and standard deviations of both groups.
  3. The effect size d is then obtained by dividing the difference between the group means by this pooled standard deviation.
  4. In meta-analysis, weights are determined by variance; fixed-effect models use only within-study variance, while random-effects models add between-study variance (T2) to the total variance.
  5. The inclusion of between-study variance in random-effects models results in different study weights and wider confidence intervals compared to fixed-effect models.

Worked example

Problem

A researcher compares a treated group (n=100, mean=50, sd=2.81) and a control group (n=100, mean=49, sd=2.81). Calculate the pooled standard deviation and the effect size d. Then, explain how the confidence interval would change if switching from a fixed-effect to a random-effects model.

  1. Calculate pooled standard deviation: Since n1=n2=100 and sd1=sd2=2.81, Swithin=(99×2.812)+(99×2.812)198=2.81.
  2. Calculate effect size d: d = (50 - 49) / 2.81 = 1 / 2.81 ≈ 0.356.
  3. Compare models: In a fixed-effect model, weights rely solely on within-study variance. In a random-effects model, between-study heterogeneity (T2) is added, altering weights and generally producing a wider confidence interval for the summary effect.

Answer S_within = 2.81, d ≈ 0.356. The random-effects model yields a wider confidence interval due to accounted heterogeneity.

Common mistakes

  • Confusing the standard deviation of the data (population variability) with the standard error of the mean (precision of the estimate).
  • Assuming that observational studies provide the same level of evidence for causality as randomized controlled trials.
  • Failing to account for between-study heterogeneity when choosing between fixed-effect and random-effects models in meta-analysis.
  • Neglecting to create necessary data fields at the start of a study, which makes retrospective data entry impossible.

Animation

Animation coming soon

The animated version of this lesson is being prepared for the web. In the Signals AI app, lessons are drawn step by step.

Code

This is a text lesson: it has no animation code yet.

ccBYNC© 2026 SignalsLab. Lesson text shared as is under CC BY-NC 4.0.