lecture 17 Flashcards
Quantitative analysis
the numerical representation
and manipulation of observations for the purpose of
describing and explaining the phenomena that those
observations reflect.
Goal of the research
To explore the relationship between learners’ socio-
demographic characteristics and their level of
participation in computer conferencing (online
discussion
what were the two theories
- Increased levels of student interaction and
engagement will lead to enhanced learning and
other educational outcomes. - Socio-demographic characteristics of students will
influence their likelihood to participate in
computer-mediated educational settings.
CALL computer conference
Informal “bulletin board.”
Dialogue on leadership-related readings.
Bilateral and small group e-mail exchange.
Orientation and debriefing for each seminar.
Work on the Issues Analysis Project
small case study
how was data gathered
Data for this study were obtained through an
unobtrusive review of the archive from the entire
CALL computer conference.
Informed consent obtained prior to start of program.
Course took place from October 1997 to May 1999.
how was the data managed
Over the course of the computer conference, regular
“housekeeping” was conducted to delete trivial
messages (such as one-word responses to existing
messages) and archive all other messages.
Messages sent by program instructors and
administrators were excluded from the analysis in
this study.
In the summer of 1999, each message that had been
archived over the course of the conference was
categorized according to
(1) the identity of the
sender; (2) the month in which it was sent to the
conference; and (3) the sub-conference to which it
was sent
this categorization was then transferred to a
spreadsheet, and the data were compiled and
analyzed using the Statistical Package for the Social
Sciences (SPSS)
Methodological limitations
Quantity of messages does not equate to quality.
Only measured the number of messages sent to
public spaces – not private e-mail communication.
Only measured the sending of messages – not the
reading of messages (lurking).
Small number of non-randomly selected participants
means that inferential statistics are questionable.
Participants: constants
Sector of employment (all worked in agriculture).
Economic resources (all paid $5,000 tuition).
Intellectual and social skills (all gained admission to a
highly competitive program – 30 / 140 admitted).
Nationality (all were Canadian).
Racialized status (all had European heritage).
Participants: socio-demographic variables
Gender (16 men; 14 women)
Age (15 under 45; 15 over 45)
Educational attainment (16 U graduates; 14 others)
Rural (24) or urban (6) residency
On-farm (16) or off-farm (14) employment
Region of residence (6 Atlantic; 7 QU & ON; 9 MB &
SK; 8 AB & BC).
Independent variables
Gender (male / female)
Age (under 45 / over 45)
Education (university graduates / others)
Residency (rural / urban)
Occupation (farmer / other)
Region (Atlantic / Central / Prairie / West)
Dependent variables
The number of messages sent by each individual
participant to the computer-mediated conference.
A ratio measure (from zero to several hundred).
how were the dependent variables interpreted
Collapsed for interpretive purposes into an ordinal
measure (low, medium, and high).
The statistically average CALL participant sent
just
under 5 messages each month to the CMC
variability of participation
Participation was highly variable, with one quarter of
learners sending less than 1.7 messages per month,
and one quarter of learners sending more than 5.7
messages per month.
Observations: variability of participation
Note the large size of the standard deviation scores
in comparison with the means, and the fact that
each median score is substantially lower than the
mean.
A relatively small number of highly active
participants contributed many messages, while most
participants contributed fewer than the mean
number of messages.
Observations: typology
Seven “high-end” users contributed nearly sixty per
cent of all messages to the computer conference,
and sent on average over twelve messages per
month.
Eight “low-end” users sent an average of only one
message per month
Reflections on the univariate analysis
How can we understand the vast differences
between levels of participation in this computer
conference?
Are such differences merely the reflection of
idiosyncratic differences in motivation, learning
styles, or receptivity to computer-conferencing as a
medium of education?
Or are there socio-demographic characteristics that
predispose certain individuals to higher or lower
rates of participation?
This leads us to bivariate analysis: to explore the
possibility that learners’ socio-demographic
characteristics explain, in part, their differing levels
of participation in the computer conference.
Bivariate analysis
Using analysis of variance (ANOVA) procedures allows
us to compare the mean level of participation by
learners with different characteristics.
Bivariate analysis: gender
Men and women had virtually identical mean levels
of participation.
There was a difference, as indicated by the standard
deviation scores, in the variability of participation of
men and women.
Men were disproportionately represented at either
the high or low end of participation, while women
were clustered more strongly in the middle.
Bivariate analysis: age
Learners’ age did not have a strong relationship
with their average level of participation.
Those learners over the age of forty-five were more
likely to be at the high or low end of the participation
spectrum, while those learners under forty-five were
more likely to be in the middle.
Bivariate analysis: education
The relationship of formal education with level of
computer conferencing participation was both
statistically significant and practically important.
Learners with university degrees sent nearly three
times the number of messages as did learners
without degrees.
The very low standard deviation score for non-
degree holders indicates that not having a university
degree was a very strong predictor of relatively low
participation in the computer conference.
All seven high-end users had university degrees.
Bivariate analysis: residence
Whether a learner lived in an urban or rural area
also had a strong and statistically significant
relationship with his or her level of participation.
The gap between rural and urban dwellers was the
largest absolute gap identified in our study.
Whether a learner lived in an urban or rural area
also had a strong and statistically significant
relationship with his or her level of participation.
The gap between rural and urban dwellers was the
largest absolute gap identified in our study.
Bivariate analysis: occupation and region
Neither occupation nor region had statistically
significant relationships with participation.
Non-farmers contributed an average of 50 messages
more than did farmers.
Those east of Manitoba contributed fewer messages
than those living in Western Canada
Reflections on the bivariate analysis
From the bivariate results, it appears as though:
— Gender and age had no impact on participation.
— Occupation and region had minor impact.
— Education and rural vs. urban residence had major impact.
However, these initial appearances may reflect the
interrelated character of different socio-economic
characteristics.
Do interrelationships between the various socio-
demographic variables in our study influence the
appearance of relationships between the individual
independent variables and learner participation?
This leads us to multivariate analysis.
The ANOVA “F” scores and statistical significance
coefficients are reported to indicate the extent to
which there is an observed difference between
participation of degree and non-degree holders within
each of the sub-groups indicated by the five other
variables
Observations on holding education constant
Gender and age still have no impact on participation.
The small impacts observed in bivariate relationships
between occupation-participation and region-
participation are reduced.
Both education level and residence in rural or urban
areas still make a meaningful difference in the
average level of learner participation.
Key outstanding question
What is the relative importance of education and rural
or urban residence when explaining the overall
variability in learner participation in this computer
conference?
Regression analysis
Despite its limited usefulness with such a small and
non-randomly-selected sample, regression analysis
provides a rudimentary estimate of the relative
strength of education and residence in urban or rural
areas as predictors of learners’ participation in the CALL
computer conference.
Regression results
The R Square value of .323 suggests that nearly one-
third of all variability in levels of participation in the
CALL computer conference can be attributed to the
influence of education and residence in rural or
urban areas.
In other words, using this regression equation to
predict the level of learners’ participation results in
predictions that are 32.3% more accurate than
simply using the mean (98.8) for predictive purposes.
regression results
As examples, a rural learner without a university
degree would be expected to have sent about 48
messages to the conference, while an urban learner
with a university degree would be expected to have
sent about 199 messages.
About two-thirds of the variability in learners’
participation in this study cannot be attributed to
the socio-demographic variables included in the
analysis.
This variability can be understood as reflecting
situational and dispositional factors, random
differences, and other structural variables that are
not included in the analysis.
conclusions
Of the six independent variables in our study, only
education and residence in urban or rural areas
were significantly related to rates of participation.
There are theoretical reasons to believe that both
higher educational attainment and living in cities
would increase such participation.
Parsimonious
means of summarizing a large amount
of data
Superficial operationalization
of the complex
concept of “participation.
quantitative analysis
Enables claims to be made about causal
relationships.
Stronger with larger samples.