How to analyze repeated measures game data with changing difficulty levels?

#1
Participants: 19 children participated in the study. Children were grouped in groups G1 (N=8) and group G2 (N=11) based on their behaviors. A pre-assessment of children behavior in the classroom is done for both the groups G1 and G2 by their teacher in a form of questionnaire.

Task: The children were trained to be calm (reduce their anxiousness/stress) in the classroom by playing a neurofeedback game controlled by the brain-computer interface headset called Neurosky. The headset gives a relaxation score (range 0-100) at the rate of 1 sample/second. This training was conducted for 22 sessions (3-5 sessions/week). In each session, the children played the game and collected 4 or 5 tokens (game points).

There are two attributes - “Relaxation Threshold (RT)” and “Minimum Time to be Relaxed (MTR)” that controls the time taken by the children to collect a game token. Say if the teacher sets RT = 40 and MTR = 4 seconds, then the child should hold their relaxation score above 40 for a total of 4 seconds to get a game token. In a session, children collects 4 or 5 game tokens.

Both the attributes RT and MTR are controlled by the teacher by remotely connecting to the child’s device. If the teacher feels that the game is too easy or difficult for the child, she can increase/decrease the difficulty of the game by change the RT and MTR values. For example, after the third token, if the teacher increases the difficulty by setting RT = 40 and MTR = 11 seconds, then to collect the 4h game token, the child needs to hold his/her relaxation above 40 for 11 seconds. Same is applicable for the 5th game token as well. Thus, there is a possibility that each game token might have different RT and MTR values (both within a session/between sessions).

After 22 sessions, a post behavior assessment was done by the teacher for both G1 and G2 to understand if the children remain more calm in the classroom after the biofeedback training.

(RQ1) Did the time taken by the children to get game tokens significantly reduce over 22 sessions for 19 participants? Is there a significant difference between the participant groups G1 and G2 in their time taken to collect game tokens?

I decided to do a Mixed ANOVA with Sessions (22) and Groups (G1, G2) as Independent variables, and time taken to collect tokens within a session as DV.

Issue I am facing:
  • Most of the sessions have 5 tokens whereas some of them have 4 tokens or 3 tokens.
  • Tokens (both within-session and across sessions) have different RT and MTR values. Thus, they have different difficulty levels.
  • There are few missing data. (1 or 0 tokens were only collected in certain sessions).
  • Children's performance could also be different. However, pre-assessment questionnaire by teacher showed that there was no significant difference between G1 and G2 in their calmness and anxiety levels.

My Query to this forum: Is there an analysis method that I can use to solve RQ1 that also address all the above listed issues?
 
Last edited:
#3
Thanks, my game data is hierarchical.

In short:
In each session, participants collect game points. Two time-varying covariates affect the time taken to collect EACH game point within a session. Participants collect 5 game points per session.
There are 22 sessions for each participants (N=19).
I am interested to see: Whether [Time taken to collect ALL game points/session] reduce over the 22 session for 19 participants and/or individual participants.

From seeing your answer, and considering the data to be hierarchical - should I perform a linear mixed model and put both my 2 time-varying covariates as random effect and as covariates?
In this case, I have one more question: my data has time taken for EACH game point within session and their 2 covariates .
However, I am interested in seeing the trend of (time taken to collect ALL tokens/session) reduce over the 22 session or not? How do i handle this?

Sorry if the question is very basic. Thanks for your help.
 
Last edited: