powerSurvEpi (version 0.1.3)

powerConLogistic.bin: Sample Size Calculation for Conditional Logistic Regression with Binary Covariate

Description

Sample Size Calculation for Conditional Logistic Regression with Binary Covariate, such as matched logistic regression or nested case-control study.

Usage

powerConLogistic.bin(
  N = NULL, 
  power = 0.8, 
  OR, 
  pE, 
  nD, 
  nH, 
  R2 = 0, 
  alpha = 0.05, 
  nTests = 1,
  OR.low = 1.01,
  OR.upp = 100
)

Arguments

N

integer. Number of sets. Each set contains nD cases and nH controls.

power

numeric. Power of the test for if the exposure variable is associated with the risk of diseases

OR

numeric. Odds ratio \(=exp(\theta)\), where \(\theta\) is the regression coefficient of the exposure variable.

pE

numeric. Population prevalence of exposure.

nD

integer. Number of cases per set.

nH

integer. Number of controls per set.

R2

numeric. Coefficient of determination of the exposure variable and other covariates

alpha

numeric. family-wise type I error rate.

nTests

integer. Number of tests.

OR.low

numeric. Lower bound of odds ratio. Only used when OR=NULL and power and N are not equal to NULL.

OR.upp

numeric. Upper bound of odds ratio. Only used when OR=NULL and power and N are not equal to NULL.

Value

If the inputs is.null(N) = TRUE and is.null(power) = FALSE, then the function returns the number N of sets.

If the inputs is.null(N) = FALSE and is.null(power) = TRUE, then the function returns the power.

Otherwise, an error message is output.

Details

The power and sample size calculation formulas are provided by Lachin (2008, Section 3.3, Formula (38))

$$power = \Phi\left( \sqrt{N c} - z_{\alpha/(2 nTests)}\right)$$ and $$N = (z_{power} + z_{\alpha/(2 nTests)})^2/ c$$ where \(\Phi\) is the cumulative distribution function of the standard normal distribution \(N(0, 1)\), \(z_{a}\) is the upper \(100 a\)-th percentile of \(N(0, 1)\), $$c = \theta^2 pE (1-pE) (1-R^2)nD*nH/(nD+nH)$$ and \(R^2\) is the coefficient of determination for linear regression linking the exposure with other covariates.

References

Lachin, JM Sample Size Evaluation for a Multiply Matched Case-Control Study Using the Score Test From a Conditional Logistic (Discrete Cox PH) Regression Model. Stat Med. 2008 27(14): 2509-2523

Examples

# NOT RUN {
# estimate power
power = powerConLogistic.bin(
  N = 59, 
  power = NULL, 
  OR = 3.5, 
  pE = 0.15, 
  nD = 1, 
  nH = 2, 
  R2 = 0, 
  alpha = 0.05, 
  nTests = 1)

print(power) # 0.80

# estimate N (number of sets)
N = powerConLogistic.bin(
  N = NULL, 
  power = 0.80, 
  OR = 3.5, 
  pE = 0.15, 
  nD = 1, 
  nH = 2, 
  R2 = 0, 
  alpha = 0.05, 
  nTests = 1)

print(ceiling(N)) # 59

# estimate OR
OR = powerConLogistic.bin(
  N = 59, 
  power = 0.80, 
  OR = NULL, 
  pE = 0.15, 
  nD = 1, 
  nH = 2, 
  R2 = 0, 
  alpha = 0.05, 
  nTests = 1,
  OR.low = 1.01,
  OR.upp = 100)

print(OR) # 3.49



# }