bivarRIpower (version 1.2)

bivarcalcn: Function to calculate necessary sample size to achieve given power for correlation parameters in bivariate linear regression model

Description

Function carries out maximum likelihood sample size calculations for one of four types of correlations in a bivariate random-intercept (RI) linear regression model discussed in Comulada and Weiss (2010): 1) Correlations between RI; 2) residuals, 3) observations measured at the same time point (concurrent); and 4) observations measured at different time points (lagged). Standard deviations for variance parameters and correlations between RI and residuals are specified by the user. Correlations between concurrent and lagged observations are calculated. Sample size is calculated for specified correlation and power under a two-sided test with a .05 alpha level. Powers for remaining three non-specified correlations are also shown.

Usage

bivarcalcn(power, powerfor, timepts, d1, d2, p, p1, s1, s2, r, r1)

Arguments

power

Power to achieve (usually at least .80)

powerfor

Correlation to base sample size calculation on. Possible entries are 'RI', Random intercepts; 'RESIDUAL', Residuals; 'YYcon', Concurrent outcome observations; or 'YYlag', Lagged outcome observations.

timepts

Number of time points

d1

Standard deviation (SD) for 1st random intercept

d2

SD for 2nd random intercept

p

Correlation between RI under null hypothesis

p1

Correlation between RI under alternative hypothesis

s1

SD for 1st residual

s2

SD for 2nd residual

r

Correlation between residual under null hypothesis

r1

Correlation between residual under alternative hypothesis

Value

Returns sample size (labled as 'clusters') and parameters specified for calculations

Details

References

Comulada WS and Weiss RE. (2010). Power calculations for correlations between bivariate longitudinal data. Statistics in Medicine. 29(27): 2811-2824.

See Also

Examples

# Example: Calculate necessary sample size to achieve 80 percent power at 5 
# percent alpha level for null and alternative hypotheses that correlation 
# between RI is 0 and .2, respectively, across 6 time points.  Other 
# covariance parameter are set as follows: Correlation between residuals = 0;
# Standard deviations: 1st RI = 1, 2nd RI = 2, 1st residual = .5, 
# 2nd residual = .75  
library(bivarRIpower)
bivarcalcn(power=.80,powerfor='RI',timepts=6,d1=1,d2=2,p=0,p1=.2,s1=.5,s2=.75,
r=0,r1=.1)