help analysins stats

spitz

Hi all,

Can someone please advice on the significance of these results in plain English please, in particular , point number 8.

thank you.

Table 1. Eigenvalues and % of total variance of the 1st and 2nd order factors and structure matrix of the 2nd order analysis.

2Scoops

A bit of context would be nice...

trish23

Lies, lies, & damn statistics....

spitz

its results from data taken from people who play on line games.

with point number 8 looking at people who felt they were effectively in another world.

thank you

2Scoops

If this is from a scientific publication, it would be a huge help if you briefly stated what the study aimed to do and included their description of their statistical analysis. Can't really interpret the results without it, tbh.

spitz

Basically they studied people who play online simulation games and role playong games.

these are online world that go on if people are in them or not.

ABSTRACT
This study introduces a psychological measurement model
for analyzing involvement and presence in digital game
context. These two constructs are both theoretically and
methodologically well developed in their own fields. The
components forming these two constructs are psychologically
relevant to our understanding of the evolvement of a
user experience in digital gaming. The measurement model
is tested with a large data (n=2182) collected from a webbased
questionnaire and laboratory experiments among PC
and console players. The results show that these two psychological
constructs can be extracted from interactive
game environments. It is also shown that involvement and
presence are different dimensions of a larger psychological
entity that describes the way the players adapt themselves
psychologically into a game- world.

..................
DISCUSSION
The aim of this study was to form a measurement model
and investigate involvement and presence in digital gaming.
Involvement and presence are both theoretically and methodologically
well developed in their own fields [7-9, 14,
18]. They are psychologically relevant in understanding the
evolvement of a user experience in human-computer interaction.
The measurement model was applied into a large
and representative data gathered among the PC and console
players in Finland.
The results show that involvement and presence dimensions
can be yielded also from a gaming context. The structure of
these dimensions is in line with previous studies [7, 9, 14].
It seems that social aspects of the game tie these two dimensions
together. The game gets its meaning and relevance
trough social interaction.
The interaction scale did not load on either dimension.
However, this cognitive evaluation of the game-world’s
interactivity (e.g., the game reacted fast to my actions) is
likely to be part of the gaming experience and should be
reconsidered in future studies. The experienced motion feelings
(e.g., I had real motion sensations) were part of the
physical presence scale, integrating perceived feelings of
action into a presence dimension [8].

spitz

RESULTS
Measurement Scales
Measurement scales were extracted in a principal axis factoring
(PFA) which included the 83 variables. A significant
Bartlett's test of sphericity and .963 Kaiser-Meyer-Olkin
(KMO) measure of sampling adequacy indicated a good
factorability of the data. Analysis yielded 14 factors with an
eigenvalue greater than 1. In order to obtain various theoretically
meaningful constructs [4], an oblique direct
Oblimin rotation (delta=0) was used. The inspection of both
the pattern and structure matrices indicated that the eight
factor solution was both theoretically the most meaningful
and easiest to interpret. An eight factor solution explained
47.04% of the total variance (see Table 1). Figure 2 shows
how the previously used scales were reorganized to form
the eight measurement scales.
Factor scores with Bartlett’s method were computed for
each measurement scale to further analyze their relationships.
The standard errors of measurement of the scales
were good ranging from 0.33-0.71. The reliability coefficients
(Cronbach’s alpha) of the extracted measurement
scales are presented in Figure 2.
The measurement scales were further analyzed in a 2nd order
PFA. A significant Bartlett's test of sphericity and .720
KMO indicated a good factorability of the data. An oblique
direct Oblimin rotation (delta=0) was also applied to the 2nd
order PFA. Interaction scale had a low communality and it
did not load on either factor. Thus, it was removed from the
model. The second order analysis revealed the latent true
scores that were in the scope of this study: involvement and
presence (see Table 1). Together these two distinct but correlated
dimensions explained 50,76% of the total variance.
Co-presence and importance loaded on both the factors.

spitz

thanks again, I didn't want to post up too much of the paper TBH. this may be tricky for you as they carried out previous work in this area and used the models they generated to come up with their conclusion.

But I'm having a hard time figuring out exactly where they are coming from.

BTW- If anyone knows of a standalone stats class please let me know.

Thanks

spitz

silverside

They claim to have identified 8 significant factors affecting (enjoyment etc) of games

1 of these is much more significant than the others, explaining 20% of variance

They've found these by statistically analysing what seems to be a big questionnaire they gave to a load of games players using "principal component analysis", the first factor "risk engagement" is calculated using 12 of the scores.

the part of the paper you've posted seems to be discusisng whether the factors are measurable and statistically significant, the more interesting thing to me is whether you can use these scores for anything....

anyway you want a basic understanding of linear algebra (eigenvalues and eigenvectors) and then "principal component analysis" as a way of finding patterns in data. possibly also of survey design.

google may be useful, or standard stats textbooks (anyone suggest?)

spitz

thank you, so would that mean that the higher the % varience, the more significant it is?

Does that meant that on Factor 1 - role engagement that because it has a percentage varience of 23.67% that it is the more reported/ experienced factor of the 8?

What I'm interested in is factor 8 - Presence, do these results show it is the lowest or the highest factor that peolpe experience when playing these games?

And I'm assuming that these % are all based/ relative to each other as that is all the guy measured, the figures refer to the factors within the study, they could very well have left out important factors.

I'm not that concerned about the rest of it, there are plenty of studies out there, but measuring presence is rare enough that I have to consider and understand this regardless if its useful or not.

thank you

silverside

yes factor 1 is the most important and explains 23 % of the result

factor 8 only explains 1% - 2% so is not very important in itself

you can keep adding factors 9,10,11 to explain more of the data - its not clear how they arrived at the decision to only have 8 factors - and (i know nothing about online gaming) its not clear that factor 8 actually means anything practical. Deciding which components are "noise" is part of the art of Principal Component Analysis.

I'm no expert but that should give you some pointers.

2Scoops

spitz wrote: »

thank you, so would that mean that the higher the % varience, the more significant it is?

Not exactly. They have constructed a model to explain the variance of something in this group of games-playing people. They then selected the 8 most relevant and easy to understand (this bit appears to be a bit subjective) factors that explain the variance and together they account for about 50% of it. All the factors are 'significant' in the statistical sense, but some explain more unique variance than others. Of the eight, #8 explains the smallest amount of unique variance once you account for the other 7, but it still adds something to the model.

spitz wrote: »

What I'm interested in is factor 8 - Presence, do these results show it is the lowest or the highest factor that peolpe experience when playing these games?

The PCA identified it as a significant predictor of variance, and of the 8-factor model chosen by the authors, it explained the lowest amount compared to the other factors. That doesn't mean it's not related or important; it's in the top 8, in this dataset.