Actor Partner Interdependence Model

Estimated by Structural Equation Modeling

CAUTION: If you decide to use information from this app in a paper/manuscript, please make sure that you acknowledge that you have used this program by citing it. Also, should you decide to copy exact text, you will need to put quotes around that material to avoid plagiarism.

Stas, L, Kenny, D. A., Mayer, A., & Loeys, T. (2018). Giving Dyadic Data Analysis Away: A User-Friendly App for Actor-Partner Interdependence Models. Personal Relationships, 25 (1), 103-119. DOI: 10.1111/pere.12230

Graphical Presentation of the Fitted Model

The full APIM figure

The two figures presented below reflect all variables included in the model, also all covariates. If not all estimates are presented clearly, please try to diminish (or enlarge) the size of your browser. The figure will automatically adapt.

Standard model:

Model with standardized parameter estimates:

The Basic APIM Figure

The plots below show the basic APIM. Covariates and other requested parameters are estimated, but for simplicity only the estimates of the actor and partner effects are presented, together with the residual nonindependence in the outcome scores. In particular, the estimates, standard error and level of significance are shown. If two independent variables are taken into account the results for both are shown in separate figures. Please note this is done to facilitate the interpretation, but both predictors are indeed included in the same model at the same time.

The Basic APIM Figure

The two plots below are identical to the ones above, but only show the basic APIM. Covariates and other requested parameters are still estimated, but for simplicity only the estimates of the actor and partner effects are presented, together with the residual nonindependence in the outcome scores. In particular, the estimates, standard error and level of significance are shown.

Standard model:

* p < .05; ** p < .01; *** p < .001

Model with standardized parameter estimates:

* p < .05; ** p < .01; *** p < .001

Actor and Partner Effects

The plots presented below are based on the fitted model. Thus, they take into account all estimated parameters.

The plot presented below is based on the fitted model. Thus, they take into account all estimated parameters.

First Independent Variable:

Second Independent Variable:

Check Normality and Outliers

In order to check the normality of the residuals, one can consult the following three plots.

Guidelines for interpretation:

- The density function of the residuals ideally looks bell-shaped.

- The boxplot should look symmetric around zero with limited number of outliers (or none at all).

- For the QQ-Plot, the dots should be situated on or close to the straight line.

Effects in Raw Data

The following plots are an exploration of actor (partner) effects, ignoring partner (actor) effects and the effects of plausible covariates.

It is important to know that these bivariate plots are strictly meant for data exploration because they do not represent pure actor (partner) effects from the APIM. Indeed, in the first plot the effect of one's own outcome on one's own predictor is not adjusted for the effect of the predictor of the partner, and hence does not strictly reflect the actor effect of the APIM. We therefore refer to these effects as 'semi-actor effects'. Similarly, the second plot rather shows the 'semi-partner effects'. To assess whether it is reasonable to assume linear effects of the independent variable on the outcome a smoother (i.e., a non-parametric best fitting curve) is added to the plots as a dotted line.

When one is interested in the estimated actor and partner effects of the fitted model, please consult the tab Figures .

First Independent Variable:

Second Independent Variable:

In the background, lavaan (i.e., an R-package for analyzing latent variables; Rosseel, 2012) is used. It is advised to save this lavaan script as a record of the analysis performed by the APIM_SEM app. This script allows to easily replicate the analysis afterwards, gives other researchers insight in your analysis and increases transparency.

The lavaan syntax can be directly copy-pasted into R for specifying a lavaan model. A step-by-step tutorial for first time R/lavaan users can be found in the tab Extra Info .

The basic APIM model

The APIM model correcting for unreliability

This model corrects for unreliability, based on the presumed reliability as entered in the 'Additional tab'. For more information on this process, please consult the paper ...

Extra Info using APIM_SEM

Background information

The Actor Partner Interdependence Model (APIM; Kashy & Kenny, 2000) is one of the most prominent models for analyzing dyadic data. Although dyadic research has become immensely popular, its statistical complexity might be a barrier for some researchers. To remedy this, a free user-friendly web application, called APIM_SEM, has been developed. This app automatically performs the statistical analyses (i.e., structural equation modeling) of both simple and more complex APIMs. The app can be used for both indistinguishable dyads (e.g., friends) and distinguishable dyads (e.g., heterosexual couples). You can select if dyad members are assumed to be distinguisable or indistinguishable in the input tab Variables .

This program allows to examine dyadic patterns, to estimate actor and partner effects of one or two predictors, and to control for covariates. Results are provided in both R output (i.e., lavaan by Rosseel, 2012) and text format, complemented by summary tables and figures. As an illustration, an example data set is included. This is data on 238 American couples of a longitudinal study of Acitelli and colleagues (Acitelli, 1997; Acitelli, Veroff, & Douvan, 2013).

The APIM_SEM app is part of a bigger project called DyadR (Kenny, 2016). DyadR is a cluster of web applications that have recently been developed to aid researchers in using and understanding both simple and more complex APIM analyses (e.g., APIM with a mediating effect).

More documentation on the Actor Partner Interdependence Model can be found in Chapter 7 in the Dyadic Data Analysis book (Kenny, Kashy, & Cook, 2006). Also the webinar on the APIM of David A. Kenny might be helpful.

In the background, lavaan (i.e., an R-package for analyzing latent variables; Rosseel, 2012) is used. It is advised to save the lavaan script as a record of the analysis that were performed by the app. This script allows to easily replicate the analysis afterwards, gives other researchers insight in your analysis and increases transparency. The lavaan script can be found in the output tab Lavaan Syntax. A step by step tutorial of lavaan can be found here. This tutorial is suitable for users who have never used R and/or lavaan. The code of the output tab Lavaan Syntax can be directly copy-pasted in R to specify the lavaan model.

As a teaching material for structural equation modeling the book of Kline (2016) can be advised.

How to use APIM_SEM

Citing this program

If you decide to use information from this app in a paper/manuscript, please make sure that you acknowledge that you have used this program by citing it correctly. Also, should you decide to copy exact text, you would need to put quotes around that material to avoid plagiarism.

Stas, L, Kenny, D. A., Mayer, A., & Loeys, T. (In press). Giving Dyadic Data Analysis Away: A User-Friendly App for Actor-Partner Interdependence Models. Personal Relationships.

Pages and colors in APIM_SEM

The APIM_SEM app uses color coding to easily explain all its features. At the top, the program is divided in three main pages, indicated with a yellow color. 'The Fitting the Actor Partner Interdependence Model' page is used to fit a model and view its output. Here, relevant information of the model can be specified in the vertical input tabs: Blue tabs are mandatory (Select Data and Variables), black tabs are optional (Additional and Download Output). The model is fitted by hitting the 'Run Analysis!' button once. After that, the program is reactive and so the output will automatically change if the input is modified. The output is displayed in the green horizontal tabs.

Data Structure

Your data set should be a dyadic data set, i.e. you should have one row for every dyad or couple. In case of indistinguishable members the distinction between two independent (dependent) variables will be arbitrary.

For more info on this type of data, the webinar of prof. David A. Kenny might be helpful. Prof. David A. Kenny also developed some Data Restructering Macros in order to easily transform your data. Of course you can also restructure your data yourself.

Note that, for the moment, only continuous and binary between-dyad covariates are accepted as covariates.

Variables

Only when a label of a distinguishable variable is provided, the observations will be treated as distinguishable members. Note that it is requested to fill in a label, and not an actual variable of the dataset.

Additional

The specified significance level will be used for the lavaan output (e.g., confidence intervals) and the text output (e.g., assess if effect is significant).

Different options for the estimation of k (i.e., ratio actor and partner effects of the same person; Kenny & Ledermann, 2010) are presented. By allowing the program to calculate the confidence interval of k by means of bootstrap sampling, dyadic patterns in the Actor-Partner Interdependence Model can be detected. In particular, the program can check whether the couple model (k = 1), actor-only (k = 0) model or the contrast model (k = -1) is most plausible. This by checking which value of k (-1, 0 1) situates in the bootstrapped confidence interval of k. Allowing bootstrap samples can be achieved by specifying so in the input tab 'Additional Options'. Please note that the monte carlo sampling technique will be very fast, this is the parametric bootstrapping method. The non-parametric bootstrapping method (i.e., regular bootstrap), may take a while to calculate.

Data

As an illustrating example the couple data gathered by Acitelli and colleagues (1997) is used. In this longitudinal study, data were collected in two waves (wave 1: March 1993 through January 1994; wave 2: 1.5 to 2 years later). The sample consists of 148 married and 90 dating American couples.

The example data set of APIM_SEM is called Acitelli. This dataset contains a selected number of variables from the study of Acitelli from the first wave. For more information about the original dataset of Acitelli and colleagues please click here . More information about the specific variables included in this app can be found in the paper of Stas, Kenny, Mayer and Loeys (In press).

References

Acitelli, L. K. (1997). Sampling couples to understand them: Mixing the theoretical with the practical. Journal of Social and Personal Relationships, 14, 243-261. doi:10.1177/0265407597142006

Acitelli, Linda K., Joseph Veroff, and Elizabeth Douvan. Couples and Well-Being Project, 1993-1995, Detroit Metropolitan Area. ICPSR22081-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2013-12-10. http://doi.org/10.3886/ICPSR22081.v1

Kashy, D. A., & Kenny, D. A. (2000). The analysis of data from dyads and groups. In H. T. Reis & C. M. Judd (Eds.), Handbook of research methods in social psychology (pp. 451-477). New York: Cambridge University Press.

Kenny, D. A. (2016). DyadR. Retrieved 06/01/2017, from http://davidakenny.net/DyadR/DyadR.htm

Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic data analysis . New York, NY: Guilford Press.

Kenny, D. A., & Ledermann, T. (2010). Detecting, measuring, and testing dyadic patterns in the actor-partner interdependence model. Journal of family psychology, 25 (3), 359-366.

Kline, R. B. (2015). Principles and practice of structural equation modeling. New York: The Guilford Press

Rosseel, Y. (2012). lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48 (2). Available at http://www.jstatsoft.org/v48/i02

Stas, L, Kenny, D. A., Mayer, A., & Loeys, T. (In press). Giving Dyadic Data Analysis Away: A User-Friendly App for Actor-Partner Interdependence Models. Personal Relationships.

About & Contact

Lara Stas studied psychology at Ghent University (Belgium). She is now a PhD student at the department of Data Analysis at the faculty of psychology. Together with Prof. David A. Kenny she wrote the APIM_SEM app. This app is part of a cluster of free user-friendly web applications for dyadic data analysis, together called DyadR (Kenny, 2016).

Although great effort has been undertaken to ensure the accuracy of results, no complete guarantee can be made about their accuracy. If you do find an error, please report it to Lara Stas (Lara.Stas@UGent.be)

Lara Stas

Ghent University

Lara.Stas@UGent.be

David A. Kenny

University of Connecticut

Actor Partner Interdependence Model

Estimated by Structural Equation Modeling

Data (*)

Load Example Data

Use Your Own Data File (**)

Reset

Please select the correct variable

Do you want to include covariates?

This covariate is:

This covariate is:

This covariate is:

Are dyad members distinguishable?

For text output, please provide the label of ..

label first role singular

label first role plural

label second role singular

label second role plural

label 1st between covariate:

label 2nd between covariate:

label 3th between covariate:

label 1st mixed/within covariate:

label 2nd mixed/within covariate:

Significance level

Center Variables

Missing data

How do you want to treat missing data?

Correct for Unreliability

independent variable?

2nd independent variable?

dependent variable?

Calculating k (*)

Outliers

Residuals more extreme than ... standard deviations are called outliers

Remove observations

Enter the observations (i.e., rows) that you want to delete from the dataset (separated by a comma)

Start a New Analysis

Download the complete output

Please select the desired output format:

Graphical Presentation of the Fitted Model

The full APIM figure

Standard model:

Model with standardized parameter estimates:

The Basic APIM Figure

The Basic APIM Figure

Standard model:

Model with standardized parameter estimates:

Actor and Partner Effects

First Independent Variable:

Second Independent Variable:

Check Normality and Outliers

Effects in Raw Data

First Independent Variable:

Second Independent Variable:

The basic APIM model

The APIM model correcting for unreliability

Extra Info using APIM_SEM

Background information

How to use APIM_SEM

Citing this program

Pages and colors in APIM_SEM

Data Structure

Variables

Additional

Data

References

About & Contact

**Data (*)**

Use Your Own Data File ()**

**Calculating k (*)**