Tables for sensitivity analyses — sens

Produces table showing the proportion of true effect sizes more extreme than q across a grid of bias parameters muB and sigB (for meas == "prop"). Alternatively, produces a table showing the minimum bias factor (for meas == "Tmin") or confounding strength (for meas == "Gmin") required to reduce to less than r the proportion of true effects more extreme than q.

sens_table(meas, q, r = seq(0.1, 0.9, 0.1), muB = NA, sigB = NA, yr, t2)

Arguments

meas	`prop`, `Tmin`, or `Gmin`
q	True effect size that is the threshold for "scientific significance"
r	For `Tmin` and `Gmin`, vector of values to which the proportion of large effect sizes is to be reduced
muB	Mean bias factor on the log scale across studies
sigB	Standard deviation of log bias factor across studies
yr	Pooled point estimate (on log scale) from confounded meta-analysis
t2	Estimated heterogeneity (tau^2) from confounded meta-analysis

Details

For meas=="Tmin" or meas=="Gmin", arguments muB and sigB can be left NA; r can also be NA as it will default to a reasonable range of proportions. Returns a data.frame whose rows are values of muB (for meas=="prop") or of r (for meas=="Tmin" or meas=="Gmin"). Its columns are values of sigB (for meas=="prop") or of q (for meas=="Tmin" or meas=="Gmin"). Tables for Gmin will display NaN for cells corresponding to Tmin<1, i.e., for which no bias is required to reduce the effects as specified.

Examples

sens_table( meas="prop", q=log(1.1), muB=c( log(1.1),
log(1.5), log(2.0) ), sigB=c(0, 0.1, 0.2),
yr=log(2.5), t2=0.1 )
#>              sigB=0  sigB=0.1  sigB=0.2
#> muB=0.095 0.9891269 0.9922163 0.9984744
#> muB=0.405 0.9055727 0.9169816 0.9550887
#> muB=0.693 0.6569836 0.6649866 0.6991222

sens_table( meas="Tmin", q=c( log(1.1), log(1.5) ),
yr=log(1.3), t2=0.1 )
#>        0.095    0.405
#> 0.1 1.772368 1.299736
#> 0.2 1.542182 1.130933
#> 0.3 1.394986 1.022989
#> 0.4 1.280396 1.000000
#> 0.5 1.181818 1.000000
#> 0.6 1.090830 1.000000
#> 0.7 1.001225 1.000000
#> 0.8 1.000000 1.000000
#> 0.9 1.000000 1.000000

# Tmin is 1 here because we already have <80% of effects
#  below log(1.1) even without any confounding
sens_table( meas="Gmin", r=0.8, q=c( log(1.1) ),
yr=log(1.3), t2=0.1 )
#>     0.095
#> 0.8     1