Moderation Regression and Simple Slopes in Lavaan

Author

Affiliation

Seungju Kim

University of Illinois, Urbana-Champaign

Introduction

Typically, when we run a linear regression in R, we will use the lm() function. And while the lm() will get most jobs done, it has several limitations:

It cannot handle missing data.
It does not have robust standard errors estimators built in.

This is in contrast to Mplus which has both of these features, with capabilities for a host of estimators (e.g., maximum likelihood with robust standard errors) and the ability to handle missing data using full information maximum likelihood estimator.

Well the great news is that you too can do this in R using the lavaan package. Finally, often times I hear that people have a hard time plotting simple slopes using lavaan, but I will be going over a very helpful function to help us!

Let’s begin!

Preparation

Like any analysis you run in R, you should always have you set up in the beginning. ## Packages

# install.packages("pacman")

If you haven’t already, I would highly recommend downloading the pacman (package manager) package. It will install and/or load a library if you haven’t automatically!

# pacman::p_load(tidyverse, lavaan, readr, here, psych, DT)

Great! Now let’s import our dataset!

Import Data

For our example today, we will be using the ACTIVE (Advanced Cognitive Training for Independent and Vital Elderly) dataset available here

# data <- read.csv(here("resource_page", "active.csv"))

Great we’ve imported our dataset! Let’s take a look at the dataset!

Descriptives

# psych::describe(data) %>%
#   round(digits = 2) %>%
#   DT::datatable()

Here we can take a good look at the desriptives of our dataset. But what we’re really interested in is running a regression model with missing data! Let’s introduce some missingness.

Introduce Missingness

# na_data <- data %>%
#   mutate(across(everything(), ~{
#     # For each value, 15% chance of becoming NA
#     ifelse(runif(n()) < 0.15, NA, .)
#   }))

Great! Let’s take a look at the data now.

Examine Missingness

# pacman::p_load(naniar) # install/load the "naniar" package
# 
# miss_var_summary(na_data)

We can see here that about 15% of all data is missing. Let’s now run our regression model!

Analysis

Multiple Regression Model

What we would normally do if we were interested in if