sem_appl

Numeric input vector for the primary outcome.

Numeric input vector for the exposure variable.

Numeric input vector for the intermediate outcome.

Numeric input vector for the observed confounding factor.

Function which uses the <code><a rd-options="lavaan" href="/link/sem?package=CIEE&version=0.1.1&to=lavaan" data-mini-rdoc="lavaan::sem">sem</a></code> function in the
<code>lavaan</code> package to fit the model
$$L = \alpha_0 + \alpha_1 \cdot X + \epsilon_1, \epsilon_1 \sim N(0,\sigma_1^2)$$
$$K = \alpha_2 + \alpha_3 \cdot X + \alpha_4 \cdot L + \epsilon_2, \epsilon_2 \sim~ N(0,\sigma_2^2)$$
$$Y = \alpha_5 + \alpha_6 \cdot K + \alpha_{XY} \cdot X + \epsilon_3, \epsilon_3 \sim N(0,\sigma_3^2)$$
in order to obtain point and standard error estimates
of the parameters
$\alpha_1, \alpha_3, \alpha_4, \alpha_6, \alpha_{XY}$
for the GLM setting.
See the vignette for more details.

In many studies across different disciplines, detailed measures of the variables of interest are available. If assumptions can be made regarding the direction of effects between the assessed variables, this has to be considered in the analysis. The functions in this package implement the novel approach CIEE (causal inference using estimating equations; Konigorski et al., 2018, <DOI:10.1002/gepi.22107>) for estimating and testing the direct effect of an exposure variable on a primary outcome, while adjusting for indirect effects of the exposure on the primary outcome through a secondary intermediate outcome and potential factors influencing the secondary outcome. The underlying directed acyclic graph (DAG) of this considered model is described in the vignette. CIEE can be applied to studies in many different fields, and it is implemented here for the analysis of a continuous primary outcome and a time-to-event primary outcome subject to censoring. CIEE uses estimating equations to obtain estimates of the direct effect and robust sandwich standard error estimates. Then, a large-sample Wald-type test statistic is computed for testing the absence of the direct effect. Additionally, standard multiple regression, regression of residuals, and the structural equation modeling approach are implemented for comparison.

Stefan Konigorski

CIEE

Estimating and Testing Direct Effects in Directed Acyclic Graphs
using Estimating Equations

Yildiz E. Yilmaz

sem_appl function

Function which uses the <code><a rd-options='lavaan' href='sem'>sem</a></code> function in the
<code>lavaan</code> package to fit the model
$$L = \alpha_0 + \alpha_1 \cdot X + \epsilon_1, \epsilon_1 \sim N(0,\sigma_1^2)$$
$$K = \alpha_2 + \alpha_3 \cdot X + \alpha_4 \cdot L + \epsilon_2, \epsilon_2 \sim~ N(0,\sigma_2^2)$$
$$Y = \alpha_5 + \alpha_6 \cdot K + \alpha_{XY} \cdot X + \epsilon_3, \epsilon_3 \sim N(0,\sigma_3^2)$$
in order to obtain point and standard error estimates
of the parameters
$\alpha_1, \alpha_3, \alpha_4, \alpha_6, \alpha_{XY}$
for the GLM setting.
See the vignette for more details.

sem_appl: Structural equation modeling approach

Description

Usage

Arguments

Value

Examples