Please visit my website for more information: http://www.comparative-education.com/. To cite this presentation, please use the following: Wiseman, A. W. (2012, April). International Cross-Time, Cross-System (XTXS) Database: Research Potential & Example. Presentation given at the CIES workshop, “Comparative and International Education Research Made Easier: How to Use Several Free Online Data Tools,” San Juan, Puerto Rico.
International Cross-Time, Cross-System (XTXS) Database: Research Potential & Example
1. International Cross-Time, Cross-
System (XTXS) Database:
Research Potential & Example
Alexander W. Wiseman
Lehigh University
aww207@lehigh.edu
2. Problems with Cross-national Data
• It is “isolated” – many datasets exist, few are
readily compatible or complementary
• It is “inaccessible” – sophisticated statistical
skills are necessary
• It is “flat” – one or two dimensions at best
3. Potential with XTXS
XTXS is XTXS is XTXS is in
“integrated” “accessible” “3D”
8. Space: Comparing Systems
• Comparing across TIME
locations, political or
geographical
boundaries, educatio
SPACE
nal systems
• Bread and butter of
comparative and
international
education research
10. Potential with XTXS
XTXS is XTXS is XTXS is in
“integrated” “accessible” “3D”
11. Sample Research Questions
1. How do health factors associate with
student achievement?
2. Impact of teacher characteristics on HIV/AIDS
curriculum implementation?
3. How does HIV/AIDS prevalence among young
adults impact the teaching profession?
– Teacher attrition due to health/infection
– Shift in association between teacher
training, education level, years experience
12. 1. How do health factors associate
with student achievement?
• Health Factors
– Health expenditure per capita (current USD), 1995-
2009
– Malnutrition prevalence, weight for age (% children
<5), 1975-2009
– Prevalence of HIV, total (% ages 15-24), 1990-2009
• Student Achievement Indicators
– PISA science literacy (15 year
olds), 2000, 2003, 2006, 2009
– TIMSS science scores (4th graders), 1995, 2003, 2007
– PIRLS reading literacy (4th graders), 2001, 2006
13. 1. How do health factors associate
with student achievement?
The cross-time element is considered first.
How well do the two sets of factors align by year?
14. 1. How do health factors associate
with student achievement?
The cross-system element is considered next.
How good is the location coverage by year for the
health factors?
15. Potential for Statistical Analyses
• Descriptive statistics (frequencies, central
tendency, variability)
• Inferential statistics (differences, relationships)
• Multilevel analysis (w/ supplementary data)
• Structural Equation Modeling (latent
variables)
• Time Series Analysis (patterns over time)
– Event History Analysis (duration or time-to-event)
16.
17.
18. 1. How do health factors associate
with student achievement?
The cross-space element is considered next.
How good is the location coverage by year for the
student achievement indicators?
19.
20.
21.
22. Potential for Statistical Analyses
• Descriptive statistics (frequencies, central
tendency, variability)
• Inferential statistics (differences, relationships)
• Multilevel analysis (w/ supplementary data)
• Structural Equation Modeling (latent
variables)
• Time Series Analysis (patterns over time)
– Event History Analysis (duration or time-to-event)
23. Potential with XTXS
XTXS is XTXS is XTXS is in
“integrated” “accessible” “3D”
24. International Cross-Time, Cross-
System (XTXS) Database:
Research Potential & Example
Alexander W. Wiseman
Lehigh University
aww207@lehigh.edu
Notes de l'éditeur
Definitely a consumer of international comparative education data, sometimes a commentator on this kind of data, too.Been using this kind of data since 1998, often TIMSS but typically supplemented by World Bank, UNESCO, and OECD data.Promising alternative to the hunt and find method of cross-national data collection. Eliminates much of the statistical inconsistencies that pulling together your own international dataset can create.A handful of current and former graduate students who are working on research with me and turning towards this data. Will discuss some of their examples.Tell you why I think XTXS has potential: empirical advantage and dimensionality.Tell you about its potential for empirical analysis of international and comparative education phenomena.
Where to look for information is always the biggest challenge when doing cross-national analysis. Bringing these data together is often a big challenge. Are the units compatible? What about the weighting when aggregating to the national/system level? And, making sure that the character and the dynamics/complexity of a phenomenon are represented by the data is always important. How much depth is there in the data?
XTXS offers international and comparative education researchers an empirical advantage because it is:IntegratedAccessible3 dimensional
It is integrated because it brings together data from many of the most frequent sources of comparative and international education data into one coherent dataset. AchievementBackground Context/CharacteristicsTrends in educational system phenomena (enrollment, resources)Etc.
XTXS is accessible because it provides an ensemble of data in two easy-to-use formats: wide datasets and long datasets.It comes downloadable as Excel, SPSS, and SASS compatible datasets, so anyone from a range of statistical software skill levels can use it.And, the format of the data allows the user to configure it in analyses in order to look across time, system, and level.
To me, the potential dimensionality of XTXS is its greatest feature.We often talk about the benefits of triangulation in empirical research. This is similar in many respects because it gives us the potential to approach any international comparative education phenomenon using cross-system, cross-time, and cross-level analyses.
XTXS provides data for the many variables at specific yearly time points for as many of the variables as possible. These time series are cross-sectional year-by-year, and the participants/data subjects are not necessarily the same from year-to-year, so this is not longitudinal or panel data.But, it does provide multiple time points, and allows us to document and investigate change over time. This is huge, since one of the biggest critiques of a lot of internationally comparative data like the TIMSS and PISA is that it is cross-sectional and so may not represent the full reality of what’s going on.The cross-time element of XTXS provides an way to track this change over time by assembling these data over several decades together in one dataset.
Cross-national education system comparisons are a staple of international comparative education research, and XTXS provides national education systems as a key unit.
A third dimension of XTXS, which requires a bit more statistical manipulation is the potential to use this data for cross-level or cross-context analysis. Conceptually, there are individual-level, classroom-level, and school-level data that are aggregated to the national level for inclusion in the dataset, but we can compare system-level trends with school-level trends as they are aggregated. For example, what do teachers report their highest level of educational attainment is versus what is the national record of teachers’ highest educational attainment. This is one example of why it is important to collect and compare data originally sourced from different levels of analysis.It is also possible to supplement XTXS with data that uses a school or individual unit without aggregation, so that a multilevel regression analysis, for example, might be used. This is especially helpful in contextualizing effects and demonstrating the impact of nested relationships among variables. So for example, we might want student-level science achievement level and gender to be used as the main dependent variable, but use the number of women in parliament as the contextual, system/national level estimation factor to tell us if there is a contextual effect of visible women leaders on girls’ science performance.
So this is the potential: XTXS offers international and comparative education researchers an empirical advantage because it is:IntegratedAccessible3 dimensionalNow, let’s take a look at some of the promise resulting from this potential.
As I mentioned at the beginning, I have several former and current graduate students that I am working with on internationally-comparative studies of health and education.And, as you can see from the 3 sample questions above, they are all related, but take slightly different directions.Explain all three, but focus on #1 for the GCC. (explain Sandi’s experience and former position with the Abu Dhabi Education Council)I’m going to take you through the process rather than show you results per se. The process shows the potential promise for the XTXS data becoming useful and used by researchers more than anything else, I think.
The first thing we had to do was to see which variables were available in XTXS for our health factor and student achievement indicators.There are many variables to choose from, but this is what we initially came up with.
Since XTXS is 3D, we considered the data’s potential for our question from each angle.Across time, we looked at the coverage and alignment of our health and student achievement indicators by year.Great coverage for the health factors, but sparse for the student achievement data.
Next we looked across systems, and again we found that for health factors there was a lot of potential.
At this point, we began to consider our statistical options. Descriptive statistics across time and system is the best place to start because it gives us a feel for what the data looks like, what some potential trends are in the data, and how to approach more sophisticated analyses. Basic inferential statistics that show significant differences between groups and relationships would be the next stop, and give us a way to build our rationale for one of the more complex analyses in response to our main question.Multilevel analyses allow us to look at nested relationships, so for example, we might be able to see if student achievement was a product of health conditions, either by virtue of the health obstacles and challenges students faced in different systems or by virtue of the health infrastructure and its reflection of a wider resource base.Structural equation model is also an option because much of what we would like to eventually investigate is not directly measured, so latent variables become much more important – such as the impact of HIV/AIDS infection rates on teacher attrition.Time series analyses could be especially relevant to us since they identify patterns in the sequence of data over time, test the impact of one or more interventions (such as teaching health-related curricula), forecast future patterns of events or compare series of different kinds of events.We thought that an event history analysis would be especially relevant because it would allow us to measure the increases or decreases in student achievement as health factors changed over time.
As I mentioned before, we are specifically interested in the Gulf Cooperation Council countries for our main question of whether student achievement and health factors are related.So, we began with a look at the changes from year-to-year in health expenditure per capita for the years available in the XTXS data.As the chart shows, overall health expenditures seem to be increasing over time in GCC countries with some specific standouts: Qatar at the high end and Oman at the low end.
To see if the differences over time were shifting as much as we thought for the GCC as a whole (and since we are leaning towards a time series/event history analysis), we plotted the lag 1 first differences in health expenditures for the GCC Mean.Overall, the trend is very much increasing in spending. So, if there is a hypothesized positive association between health expenditure and student achievement, we should see an increase in student achievement as well when we get to that stage.
The problem is that for student achievement the coverage by year and system is much less frequent, and for the GCC countries it is even less.
The good news is that we can supplement this data with specific achievement item data from the Trends in International Mathematics and Science study science assessments for 8th grade equivalent students.These items help us not only get at the health and student achievement relationship, but they show us students’ specific knowledge related to health.