Deck 8 Flashcards
1
Q
How to read an article?
A
- What is the research objective?
- What is the general research question?
- What is the research hypothesis.
- What is the path diagram for key constructs in this research?
- How are the key constructs operationalized,
- What can you say about the validity of the research? (requires expert knowledge)
- What is the knowledge gap? Does it correspond with the internal objective?
2
Q
What is longitudinal study design?
A
- when you research things over time.
3
Q
What are the three types of Longitudinal studies?
A
- Trend study - everytime a different sample from the same population is measured.
- Panel study - the same sample from the population is measured repeatedly over time.
- Cohort study - cohorts are groups of units that have one specific characteristic in common (birth year, year of using irrigation system, year of graduation, etc.). Cohort can be both trend on a panel.
4
Q
What exactly is trend study?
A
- Different samples are drawn from the same population at two or more points in time: repeated cross - sectional design. Ideally, an equivalent measurement instrument is used. In practice, in loop trend studies (eg. CPI, GDP often alterations are made. Trend study is appropriate if you want to map changes at the level of a population over a period and/ or forecast trends, Research questions are about developments in time. Characteristics of trend studies: you can find the aggregate trend, but not individual developments; all limitations of cross - sectional design still apply (eg., regarding spurious relationships).
5
Q
Describe a reason for having an equivalent measurement?
A
- If your instruments no longer are appropriate or valid you may want to change them over time. If the ecology of a forest changes rapidly and proxies or indicators used are suddenly not reflective of my underlying construct anymore, I should change them in ways that again captures the underlying construct in a changed reality. (e.g. health at various life stapes gf a butterfly).
6
Q
What features of a trend study limit the contribution it can make to a causal question?
A
- Each time we take a different sample. So we cannot track changes in the same group. There may in fact be steep positive changes for half of the population and steep declines in the other half and complete stability on the aggregate level.
7
Q
What exactly is panel study design?
A
- Research questions are about developments in time of research units. Ideally, one sample from the population is measured repeatedly over time. In practice because of drop-out of research units, the remainder of the initial sample is usually replaced by new research units (“ refreshment samples”) and/or the initial sample is enlarged. Panel is at the level of units while trend is at the level of population.
8
Q
What are the major problems in panels?
A
- Selective drop-out (attrition, mortality). Drop out is very rarely random, refreshments likely do not match those who dropped out so over time the representivity of the sample degrades. So the main issues are:
- imperfect matching. we tend to not know about the real reason for drop out and only know so much about our units, so we may replace them with units that are only superficially similar.
- new units, even if perfectly matched on relevant characteristics, cannot be strictly compared to the rest of the sample because they have not yet been conditioned by testing, while the rest of the sample have already been tested before.
- Conditioning. By repeatedly measuring units, we may change the value of what we are trying to measure. So we confound the measurement of true change with the effects of our measurement/instrument.
9
Q
What is a single cohort design vs. multiple cohort?
A
- A simple cohort is studied over an extended period. E.g. 2014/2015 Msc Students are surveyed every 5 years from…
- Multiple cohorts are studied over an extended period, e.g. 2014/2015 MSc students, first year 2014/2015 Bsc students etc.
Both examples are panel.
10
Q
What are the three types of time - related effects to consider for longitudinal study?
A
- Age effects (maturation of the research unit).
- Period effects (history): changes over time in the variable of interest due to an event that affects units of all ages.
- Cohort effects. differences between cohort groups.
11
Q
When sampling in time what things to keep in mind?
A
- How big should the intervals be?
- Should they be equally spaced?
- What function does “ time “ serve?
- How many time points do we need?
All of these depend on the nature of the research question and characteristics of the population (should be informed by substansive knowledge about your context and population.)
12
Q
What is reliability in longitudinal design?
A
- It is precision of measurement at a given measurement occasion. (same as in cross - sectional) + Additions:
- precision/ random error over repeated measurement occasions. It is an issue when tracking intra - individual change with few time points,
- issues that are related to instances when precision of the same instrument differs in time.
13
Q
What is measurement validity in longitudinal design?
A
- Same as in cross sectional it is about the coverage of concept and accuracy of measurement
- It needs to have equivalent coverage accuracy over time.
- And it needs to track changes in systematic error of the same instrument over time.
14
Q
What is internal validity in longitudinal design?
A
- It is the same as in cross - sectional: confounder variables need to be taken care of + needs to deal with changing confounders over time.
- Needs to deal with design and model misspecification + dealing with it over time.
- Conditioning (in panel): needs to deal with testing effects and make sure that true change is not confounded (when we induce change by testing).
15
Q
What is external validity in longitudinal design?
A
- Same as in cross sectional: generalisation from sample to target population, to other settings, times, populations + generalisation of time points to time domain under study.
- Problems of non - response (selective non - response in cohort or attrition/ mortality/ dropout in panel studies) become dynamic: i.e. people may respond at first, but then dropout, response may be selective by cohort, non - response and sampling issues may be different by time point for trend studies.
