Instrumentation & Measurement Magazine 24-6 - 33

Table 1 - Variables, measures, variable operationalizations, and data-collection methods (continued)
Variables and measures
Type
Name
Subject variable
Age
Gender
Description
Participant
identification
(id)
Participant age
Participant gender
Played with
family
Subject (level2)
Cognitive
impairment
Whether the
participant
played the
mobile game in
the company of
a loved one
Whether the
participant
had a cognitive
impairment
Game played
The mobile game
randomly
assigned to the
participant
Previous
experience
Whether the
participant
had previous
experience
using computer
games
accounting for 76.64% of the variance in the data. Any covariate
with an item loading with a factor lower than 0.6 was
removed from the factor. The screen plot inspection eigenvalues>1
and the items and their loadings on the six factors are
available on request. The six factors with their respective items
were: participants' performance (Game level), the type of mobile
game played (Type of game), participants' cognitive status
(Cognitive impairment), participants' affect (Affect), whether
participants had previous experience using computer games
(Prev Exp Comp Games), and disturbances during the sessions
(Disturbances). In the context of this study, the factor items are
our covariates. In this case, there was one item or covariate per
factor. In sum, six covariates for the 12 eligible candidates (see
Tables 1 and 2) were included in the MLMM.
Predictors of Engagement
Table 2 shows the results obtained by fitting four incrementally
MLMM, namely A, B, C, D, and E, and (3)-(7) show
September 2021
Random factor
Fixed factor
covariate
Fixed factor
covariate
Type
Measurement and
data-collection
method
Participant selfreport
intake
form
Participant's SelfReport
intake
form
Participant's selfreport
intake
form
Fixed factor
covariate
Session data
collection form
Operationalization
[0-14]
[0-n]
Female=1, Male=0
Yes=1
No=0
Fixed factor
covariate
Fixed factor
covariate
MoCA assessment
tool [2]
Vibrant Minds
platform
Yes=1, No=0
Word-Search=1
Bejeweled=2
Mahjong=3
Fixed factor
covariate
Participant's selfreport
intake
form
Yes=1
No=0
the MLMM specification of each model. MLMM A (the Null
model) indicates that there was a significant variance in engagement
scores between older adults, regardless of time.
This is quantified by the ICC (between the groups). The estimates
suggest that 56% of the variability in engagement
scores was between older adults, whereas the remaining
44% was within older adults throughout the sessions. The
estimates of MLMM A also show that the grand mean for
engagement scores at the participant level was 26.68 (see
the Descriptive Statistics section). The high variability percentage
between the groups support the use of a MLMM
as a feasible modelling strategy. MLMM B, C, and D show
lower values of -2 RE/ML log-likelihood, AIC, and BIC.
As expected, Model E showed higher values of the AIC,
BIC, and -2 RE/ML log-likelihood in comparison to Model
D. Therefore, in using " the smaller is better " criterion for
goodness of fit, the interpretations of our results are based
on MLMM D.
IEEE Instrumentation & Measurement Magazine
33
Instrumentation & Measurement Magazine 24-6

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-6
