Power Analysis with Plain English Interpretation
Usage
power_interpret(
test,
effect_size,
n = NULL,
alpha = 0.05,
power = 0.8,
n_groups = 2,
n_predictors = 1
)Arguments
- test
The statistical test. One of "ttest.one", "ttest.two", "ttest.paired", "anova", "correlation", "chisq", "regression".
- effect_size
The expected effect size. Use Cohen's conventions: small = 0.2, medium = 0.5, large = 0.8 for t-tests; small = 0.10, medium = 0.25, large = 0.40 for ANOVA; small = 0.10, medium = 0.30, large = 0.50 for correlation.
- n
Sample size per group. If provided, calculates achieved power. If NULL, calculates required sample size.
- alpha
Significance level. Default 0.05.
- power
Desired power level. Default 0.80.
- n_groups
Number of groups (for ANOVA only). Default 2.
- n_predictors
Number of predictors (for regression only). Default 1.
Value
An object of class statease_power containing
power analysis results and interpretation. Use print()
to display the formatted report.
Examples
# Calculate required sample size for independent t-test
result <- power_interpret("ttest.two", effect_size = 0.5)
print(result)
#>
#> -- statease Power Analysis Report
#> Test : Independent Samples T-Test
#> Mode : Calculate required sample size
#> -----------------------------------------------------------------
#> Effect size : 0.500 (medium)
#> Alpha : 0.05
#> Desired power: 0.80 (80%)
#> Required n : 64
#> Total N : 128 (2 groups x 64)
#> -----------------------------------------------------------------
#> Interpretation:
#> To detect a medium effect (effect size = 0.50) with 80% power at alpha = 0.05, you need at least 64 participants per group (128 total for 2 groups).
#>
#> NOTE: Power analysis results are estimates based on assumptions about effect size, alpha, and power. Actual results may differ depending on the true effect size in the population.
#> NOTE: Effect sizes should ideally be based on previous research, pilot studies, or theoretically justified values, not chosen arbitrarily to reduce required sample size.
#> NOTE: A power of 0.80 is a conventional minimum. In high stakes research such as clinical trials, a higher power of 0.90 or 0.95 is often recommended.
#> NOTE: Power analysis assumes that the chosen statistical test and its assumptions are appropriate for the data.
#> -----------------------------------------------------------------
#>
# Calculate achieved power for given sample size
result2 <- power_interpret("ttest.two", effect_size = 0.5, n = 30)
print(result2)
#>
#> -- statease Power Analysis Report
#> Test : Independent Samples T-Test
#> Mode : Calculate achieved power
#> -----------------------------------------------------------------
#> Effect size : 0.500 (medium)
#> Alpha : 0.05
#> Sample size : 30
#> Achieved power: 0.4779 (47.8%)
#> -----------------------------------------------------------------
#> Interpretation:
#> With n = 30 and a medium effect size (0.50), the achieved power is 47.8%.
#>
#> WARNING: Power is less than 0.80. The study may be underpowered, there is a meaningful risk of failing to detect a true effect (Type II error). Consider increasing the sample size.
#> WARNING: Post-hoc (observed) power calculations should be interpreted cautiously, as they are largely determined by the observed p-value and effect size.
#>
#> NOTE: Power analysis results are estimates based on assumptions about effect size, alpha, and power. Actual results may differ depending on the true effect size in the population.
#> NOTE: Effect sizes should ideally be based on previous research, pilot studies, or theoretically justified values, not chosen arbitrarily to reduce required sample size.
#> NOTE: A power of 0.80 is a conventional minimum. In high stakes research such as clinical trials, a higher power of 0.90 or 0.95 is often recommended.
#> NOTE: Power analysis assumes that the chosen statistical test and its assumptions are appropriate for the data.
#> -----------------------------------------------------------------
#>