Sample Size Calculation

Sample Size Calculation

Sample Size Calculation

On this page

Sample Size Calculation - Sizing Significance

Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.
ErrorDescriptionSymbolProbabilityImplication for Sample Size
Type IRejecting a true null hypothesis (H₀) (False Positive)$\alpha## Sample Size Calculation - Sizing Significance
Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

| $\alpha## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

    | Lower $\alpha## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

(e.g., $\mathbf{0.01}## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

vs $\mathbf{0.05}## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

) → ↑ sample size | | Type II | Failing to reject a false H₀ (False Negative) | $\beta## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

| $\beta## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

    | Lower $\beta## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

(i.e., higher Power $1-\beta## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

, e.g., $\mathbf{90%}## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

vs $\mathbf{80%}## Sample Size Calculation - Sizing Significance Crucial for study precision, validity, ethical conduct, & optimal resource use.

  • Core Concepts:

    • Population: Entire group of interest.
    • Sample: Subset of population studied to make inferences.
    • Parameter: True population value (e.g., population mean $\mu$).
    • Statistic: Sample-based estimate of a parameter (e.g., sample mean $\bar{x}$).
    • Sampling Error: Inevitable difference between statistic & parameter; ↓ with ↑ sample size.

  • Key Determinants of Sample Size:

    • Significance level ($\alpha$): Probability of Type I error. Commonly $\mathbf{0.05}$.
    • Power ($1-\beta$): Probability of detecting a true effect if it exists. Aim for $\mathbf{80%}$ or higher. ($\beta$ is probability of Type II error).
    • Effect Size: Smallest difference considered clinically important. Larger effect size → smaller sample needed.
    • Variability: Population variance (e.g., Standard Deviation for continuous data, proportion for categorical). Higher variability → larger sample needed.
    • Confidence Interval (CI) width: Desired precision of the estimate.

) → ↑ sample size |> ⭐ Decreasing the significance level (e.g., from 0.05 to 0.01) or increasing the power (e.g., from 80% to 90%) will increase the required sample size.

![Type I and Type II Errors](hypothesis testing errors)

Sample Size Calculation - Formula Frenzy

  • Estimating Population Proportion (Qualitative Data):

    • Formula: $n = (Z_{1-\alpha/2}^2 * p * (1-p)) / d^2$
      • p: Expected proportion in the population.
      • d: Desired absolute precision or margin of error.
      • $Z_{1-\alpha/2}$: Z-value corresponding to the chosen confidence level.

    ⭐ When estimating a population proportion and the expected proportion (p) is unknown, use $p = \textbf{0.5}$ to get the maximum (most conservative) sample size.

  • Estimating Population Mean (Quantitative Data):

    • Formula: $n = (Z_{1-\alpha/2}^2 * \sigma^2) / d^2$
      • $\sigma$: Population standard deviation.
      • d: Desired absolute precision or margin of error.
      • $Z_{1-\alpha/2}$: Z-value for the confidence level.

  • Comparing Two Groups (Proportions or Means):

    • Key factors influencing sample size include:
      • Significance level ($\alpha$, determines $Z_{1-\alpha/2}$).
      • Statistical power ($1-\beta$, determines $Z_{1-\beta}$).
      • Magnitude of the difference to be detected between groups.
      • Population variance ($\sigma^2$) for means, or proportions ($p_1, p_2$) for proportions.

  • Key Influences on Sample Size (n):

    • n ↑ as: Confidence level ↑ (e.g., 99% vs 95%), Power ↑ (e.g., 90% vs 80%), Data variability ($\sigma$, or p(1-p)) ↑.
    • n ↑ as: Effect size (difference to detect) ↓ (smaller differences require larger samples).
    • n ↓ as: Desired precision d ↑ (i.e., a wider margin of error is accepted).

  • Common Z-values for Confidence Intervals (CI):

    • 90% CI: $Z = \textbf{1.645}$
    • 95% CI: $Z = \textbf{1.96}$ (Most commonly used)
    • 99% CI: $Z = \textbf{2.58}$
    • 📌 Remember: 1.645 (90%), 1.96 (95%), 2.58 (99%)!

Sample Size Calculation - Smart Sizing

  • Finite Population Correction (FPC): For small populations (e.g., sample >5% of total). $n' = n / (1 + (n-1)/N)$.
  • Design Effect (DEFF): For cluster sampling. $n_{cluster} = n_{srs} * DEFF$. DEFF typically $\textbf{>1}$.

    ⭐ The design effect (DEFF) for cluster randomized trials is almost always greater than 1, meaning cluster sampling requires a larger sample size than simple random sampling for the same power.

  • Non-Response/Attrition Adjustment: Increase calculated sample size. $n_{adjusted} = n / (1 - \text{dropout rate})$. E.g., for 20% dropout, divide $n$ by 0.8.
  • Diagnostic Test Studies: Sample size calculations focus on achieving desired precision for sensitivity, specificity, PPV, or NPV.
  • Pilot Studies: Rule of thumb: 10-20% of the estimated main study sample size, or 30-50 participants. Helps refine methods and estimate parameters like variance or effect size for the main study's sample size calculation.

High-Yield Points - ⚡ Biggest Takeaways

  • Sample size increases with higher power, higher confidence level, smaller effect size, and greater variability.
  • Power (1-β), typically 80%, is the probability of detecting a true effect.
  • Significance level (α), usually 0.05, is the chance of a Type I error.
  • Formulas differ for proportions (using p, d) and means (using σ, d).
  • Smaller effect size or greater variability (SD) demands a larger sample.
  • Cluster sampling requires increased sample size due to design effect.

Practice Questions: Sample Size Calculation

Test your understanding with these related questions

Specificity of a diagnostic test is defined as:

1 of 5

Flashcards: Sample Size Calculation

1/10

Statistical power of a test is calculated as _____

TAP TO REVEAL ANSWER

Statistical power of a test is calculated as _____

1-(Type II error)

←→browseSpaceflip

Enjoying this lesson?

Get full access to all lessons, practice questions, and more.

Start Your Free Trial