Some more embodiment analyses

Here are some more (explorative) analyses from the embodiment data used the Embodiment in character-based video games.

I collected also workload data using raw Nasa TLX when gathering data for EFA and CFA, but then I did not use workload data in analyses. My assumption was that workload would correlate with the embodiment, but did not look at this.

The first run of linear mixed models (using lme4::lmer)  showed issues with model assumptions and I dropped out games that have less than eight answers. The linear mixed model results are shown in Fig1 and they indicate a significant effect. However, even after dropping the games the residuals are not normally distributed and variance of random effects is practically zero. Hence, I run the model using robust lmm which give similar results: rlmer’s workload estimate is 0.295 vs lmer’s 0.301). As the impact of the game is low (the random effect is small), Pearson correlation is a feasible approach for calculating the relation between variables: correlation: r=0.26, p=0.0023, n=138. Correlation is again in line with the mixed model results.

So there seems to be a small relationship between experienced workload and embodiment. This would be natural, as when workload increases, the players are not able to focus their game controllers.

Fig 1. Model: embodiment ~ workload + (1|Game)

The study 4 data of the embodiment study allows looking the relation between completion times and the sense of presence. Unlike, the embodiment (cf. study 3), the presence does not seem to relate to the embodiment (see fig 2). Notably, this model has autocorrelation and I am unsure how to reduce that.

Fig 2. Model: Presence ~ Trial * log(Time) + (1|id)

The study 4 data of the embodiment study allows looking the relation between completion times and the embodiment. The similar relation between completion times and embodiment is seen in the static 3rd person camera condition which is the same than in study 3 (see fig 2). The embodiment and completion times does not relate to other conditions. However, this model has autocorrelation and I am unsure how to reduce that and that autocorrelation can have an impact on the results.

If the results can be trusted, the results are in line with the study 3 in the embodiment paper. Notably, the embodiment relation to completion times was not present in the 1st person and traditional 3rd person camera conditions.

Fig 3. Model: Embodiment ~ Trial * log(Time)


Published by lankoski

Petri Lankoski, D.Arts, is a Associate Professor in Game Studies at the school of Communication, Media and IT at the Södertörn University, Sweden. His research focuses on game design, game characters, role-playing, and playing experience. Petri has been concentrating on single-player video games but researched also (multi-player) pnp and live-action role-playing games. This blog focuses on his research on games and related things.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: