TVEM requires coverage of the time axis
Because TVEM estimates associations between variables across continuous time, researchers need to have adequate data on the outcome across the entire span of time they are interested in. For example, in the longitudinal panel data example, data on sexual behavior was collected at all ages from 14-32. If the study design had, instead, sampled every participant at ages 14, 22, and 32, then there would be gaps in the coverage, and TVEM might not be an appropriate method.
Coverage can be achieved through a large number of participants or waves
This coverage of the time axis can be achieved in several types of data. When intensive longitudinal data, such as ecological momentary assessment, is used, each person may have many occasions of data across the time axis. On the other extreme, cross-sectional data can be used in TVEM if there are enough people in the study assessed at different times. For example, if you had data from people at every age from ages 18-65, you could run a TVEM covering this age period.
Examples: EMA and longitudinal panel data
Our first TVEM example uses ecological momentary assessments (EMA). These data were collected at five random times each day on a cell phone for two weeks after a smoking quit attempt. By overlaying all participants’ data measured in minutes from when they quit smoking, these data provide adequate coverage over time.
Our second TVEM example uses longitudinal panel data from a cohort-sequential design. Participants completed four assessments during adolescence and young adulthood. However, participants were recruited when they were in grades 7-12. Because of this staggered entry into the study, there is adequate coverage across age (measured in months old) when combining across both people and assessments.
Any data type with adequate coverage of the time axis can be used in TVEM
Ecological momentary assessment (EMA), panel studies with frequent assessments or a cohort-sequential design, or cross-sectional data with a range of ages can be used. However some designs, like panel data that follows a birth cohort at set age intervals, may be more problematic due to gaps in the of coverage of time.
TVEM requires a meaningful “zero” on the time axis
In the smoking cessation study, everyone’s data was collected begining at the moment they quit smoking; this enabled the researchers to examine how effects differed across time. If a study collected EMA during randomly selected weeks in a subject’s normal life, there may not be a meaningful zero point, and TVEM may not be appropriate. In the panel study, subjects’ ages are on the time axis, and so birth becomes the zero point for measurement. Whatever the data type, the data must be organized so that the starting or zero point is the same for each subject.
Researchers need to be careful about interpretations
As in any study of development, researchers need to think carefully about whether differences observed are due to age or whether they could be caused by cohort effects. This is particularly problematic when cross-sectional or panel data with a wide age range and few measurement occasions is used.