Pool a scalar estimate across multiply imputed data sets

Rubin's rules: Qbar = mean(Q_l), Ubar = mean(U_l), B = var(Q_l), T = Ubar + (1 + 1/m) B, with the relative increase in variance, the proportion of variance attributable to missing data lambda, the fraction of missing information, and the degrees of freedom (Barnard-Rubin adjusted when nu_com is given).

Usage

morie_miefcl(estimates, variances, nu_com = NULL)

Arguments

estimates

Numeric vector of complete-data estimates.

variances

Numeric vector of complete-data variances.

nu_com

Optional complete-data degrees of freedom (n - k).

Value

A list with elements estimate, se, t, ubar, b, m, riv, lambda_, fmi, df, method.

References

Rubin, D. B. (1987). Multiple Imputation for Nonresponse in Surveys. Wiley. van Buuren, S. (2018). Flexible Imputation of Missing Data, 2nd ed., Chapman & Hall/CRC. Barnard, J. and Rubin, D. B. (1999). Biometrika, 86, 948-955.

Examples

V <- c(1, 2, 3, 4, 5, 6, 7, 8)
morie_miefcl(V, V)
#> $estimate
#> [1] 4.5
#> 
#> $se
#> [1] 3.354102
#> 
#> $t
#> [1] 11.25
#> 
#> $ubar
#> [1] 4.5
#> 
#> $b
#> [1] 6
#> 
#> $m
#> [1] 8
#> 
#> $riv
#> [1] 1.5
#> 
#> $lambda_
#> [1] 0.6
#> 
#> $fmi
#> [1] 0.6356436
#> 
#> $df
#> [1] 19.44444
#> 
#> $method
#> [1] "Rubin's rules (Rubin 1987; van Buuren 2018 Sec. 2.3)"
#>