Combines per-variant ratio estimates of a causal effect into a single inverse-variance weighted estimate under one of three meta-analysis models, using either first- or second-order delta-method weights on the ratio variance (Burgess & Bowden 2015).
Usage
morie_mtr2sx(
beta_x,
se_x,
beta_y,
se_y,
model = "multiplicative",
weights = "first_order",
theta = 0
)Arguments
- beta_x
Numeric vector of variant–exposure associations.
- se_x
Numeric vector of standard errors of
beta_x.- beta_y
Numeric vector of variant–outcome associations.
- se_y
Numeric vector of standard errors of
beta_y.- model
One of
"multiplicative"(default),"fixed"or"additive".- weights
One of
"first_order"(default) or"second_order".- theta
Correlation between the two association estimates, used only when
weights = "second_order"; zero in a genuine two-sample design.
Value
A named list with elements estimate, se,
z, p_value, ci (length-2 numeric),
ratio_estimates, variances, weights_used,
model, phi_multiplicative, tau2,
Q, df, I2, se_fixed,
regression_estimate, regression_se_fixed,
n_variants and method.
References
Burgess, S. & Bowden, J. (2015). Integrating summarized data from multiple genetic variants in Mendelian randomization. arXiv:1512.04486 [stat.AP].
Examples
morie_mtr2sx(beta_x = c(1, 2, 3, 4, 5, 6, 7, 8), se_x = c(1, 2, 3, 4, 5, 6, 7, 8),
beta_y = c(1, 2, 3, 4, 5, 6, 7, 8), se_y = c(1, 2, 3, 4, 5, 6, 7, 8))
#> $estimate
#> [1] 1
#>
#> $se
#> [1] 0.3535534
#>
#> $z
#> [1] 2.828427
#>
#> $p_value
#> [1] 0.004677735
#>
#> $ci
#> [1] 0.3070354 1.6929646
#>
#> $ratio_estimates
#> [1] 1 1 1 1 1 1 1 1
#>
#> $variances
#> [1] 1 1 1 1 1 1 1 1
#>
#> $weights_used
#> [1] "first_order"
#>
#> $model
#> [1] "multiplicative"
#>
#> $phi_multiplicative
#> [1] 1
#>
#> $tau2
#> [1] 0
#>
#> $Q
#> [1] 0
#>
#> $df
#> [1] 7
#>
#> $I2
#> [1] 0
#>
#> $se_fixed
#> [1] 0.3535534
#>
#> $regression_estimate
#> [1] 1
#>
#> $regression_se_fixed
#> [1] 0.3535534
#>
#> $n_variants
#> [1] 8
#>
#> $method
#> [1] "inverse-variance weighted MR (multiplicative model, first_order weights); Burgess & Bowden (2015) Sec. 2"
#>