How to measure- effect of additional categorical DV on repeated m. ANOVA?

#1
Hello,

I am helping a colleague analyze data from a user study. I am a little unclear how to deal with one of her measures. I would greatly appreciate any help!

Description:

Independent variable:
-interface modality:
-audio (A), visual (V), haptic (H), control (C- no interface)

Main dependent variable:
-breathing rate (continuous, ratio)

Additional measure
-participants reported which of the three experimental interfaces (one of A, V or H) was their favorite.
Participants did NOT rank the other two interfaces, and were not allowed to choose the control.
E.g., participant 1 might say "My favorite was visual". But we don't know whether participant 1 preferred haptic to audio, or whether, given the option, they would have said "I didn't like any of them" (preferred control).


Design:
repeated measures. each participant experienced each interface, and the control. We obtained a breathing rate measurement for each interface. (order counterbalanced across participants)

The research questions:
1. Does breathing rate depend on interface?
2. Does participants' preferences "play a role"?

(1) warrants a one-way repeated measures ANOVA (breathing rate by interface). But I am not sure how best to assess (2). The hypothesis is that participants' -preferred interface- will influence their breathing rate, whether it is haptic, visual, audio, etc. (I.e., the effect of interface modality overall should be weak, because the favorite interface varies. therefore to observe an effect of the interface- over the control- you need to factor in preference).

ATM, I've done a cursory "check" of the hypothesis by running two independent anovas. The first is just (1). For the second, I created three categories- "Preferred", "Not preferred" and "control". I then averaged each participants scores for the two non-preferred interfaces, and ran an ANOVA on the breathing rate using the preference category as a factor.

I know that this is not valid (modality and preference are not independent), and also does not directly COMPARE the relative influences of interface modality and participants' preferences. But I'm not sure what the proper test is.

I'm wondering does anyone have some thoughts on which test is appropriate, and how to model the preference variable?

Thanks!
Best,
~Emily
 
Last edited:

Karabiner

TS Contributor
#2
Why not a mixed Anova with modality as within-subjects-factor and preferred modality as beween-subjects (grouping) factor?

With kind regards

K.
 
#3
Thanks! I had considered that, but the problem is that I think I need to compare each subjects scores with their preferred modality to those of the modality they did not prefer, and the control (and predict that participants are better with the modality they prefer- ignoring what modality it is- versus the non-preferred). Otherwise I would be comparing the group of participants who preferred audio to the group the preferred visual and to that which preferred haptic, but I think that might be answering a different question then "did participants generally do better with the modality that they preferred then those they did not?"
Please let me know if I misunderstood your suggestion. I appreciate your help very much!
~Emily
 

Karabiner

TS Contributor
#4
As far als i can see, you will be able to analyse whether the groups respond differently to the different conditions (interaction effect) and whether there is a main effect of conditions. You can perform further analyses if main effect(s) and/or the interaction is statistically significant.

With kind regards

K.
 
#5
Ah, I see! So you think a 2-way mixed ANOVA modelling feedback modality as the within-subjects factor and what-they-preferred as the between-subjects factor; checking the interaction between them would directly test the question about the moderating influence of preference on modality.
Thanks!