Skip to contents

dySim

The goal of dySim is to simplify the process of running a variety of simulations for the purpose of evaluating methods of dyadic data analysis. dySim piggybacks heavily on the lavaan, simsem, and dySEM packages.

Installation

You can install the development version of dySim from GitHub with:

# install.packages("devtools")
devtools::install_github("jsakaluk/dySim")

Collaboration and Troubleshooting

Please submit any problems or feature requests via the dySim issues page.

If you are interested in collaborating on the development of dySim, please contact Dr. Sakaluk.

General Workflow

  1. Identify population model/data generating process from which you wish to draw simulated random samples

  2. Save a popModList with the arguments appropriate for the population model you have chosen in Step 1

  3. Run runDySim() with your chosen model in Step 1 and its popModList, as well as indicating your preferences for features of your analysis of each simulated random sample, and what your analytic target is for the simulation. Behind the scenes, dySim will specify your population model’s parameters, feed it to simsem for generating random samples from this population, which will be analyzed using a combination of dySEM (for sample model scripting) and lavaan (for sample model fitting).

  4. Summarize, analyze, and vizualize the data frame returned by runDySim() in order to make sense of your simulations!

Supported Population Models/Data Generating Processes

  • Latent Actor-Partner Interdependence Model
  • Latent Common Fate Model

Each kind of parameter in each model accepts a type = argument (specified in the popModList) in order to control what kinds of values are specified.

Supported Sample Models

  • Latent Actor-Partner Interdependence Model
  • Latent Common Fate Model
  • Observed Actor-Partner Interdependence Model
  • Observed Common Fate Model

Example

This is a basic example which shows you how to solve a common problem:

library(dySim)

#set population parameters for model in a list with features
#needed for a latent APIM data generating mechanism
popModList.apim <- list(nItemsX = 5, #five items for measure of latent x 
                   loadValuesX_A = "moderate", #A's X-loadings are moderate
                   loadValuesX_B = "moderate", #B's X-loadings are moderate
                   residCorrValuesX = "moderate", #residual correlations between A's and B'x X-indicators
                   iccX = "weak", #intraclass correlation between A and B's latent-Xs
                   nItemsY = 5, #five items for measure of latent y 
                   loadValuesY_A = "weak", #A's Y-loadings are moderate
                   loadValuesY_B = "mixed", #B's X-loadings are moderate
                   residCorrValuesY = "mixed", #residual correlations between A's and B'x X-indicators
                   iccY = "strong",# residual intraclass correlation between A and B's latent-Xs
                   actorA = "moderate", #actor effect for A 
                   actorB = "very strong", #actor effect for B
                   partnerA = "weak",  #partner effect for A
                   partnerB = "weak") #partner effect for B

#run simulation and save combined parameter table from all 5000 sims in a df
lapim.lapim.100 <- runDySim(seed = 123,#set random seed for reproducibility
                    popMod = "L-APIM", #generate from latent APIM pop. mod
                    popModList = popModList.apim, #supply list of model parameters
                    sampSize = 100, # n for each random sample
                    sampMod = "APIM", # what model to fit in each random sample
                    sampModType = "latent", #whether to use "latent" or "observed" version of model to fit in each random sample
                    nSims = 5000, #number of random samples to draw/fit model in
                    output = "paramTable") #what kind of output to extract from each sample's analysis