Temporal design and its consequences
This article builds on Hamaker (2023).
This article is about the temporal design of your study, and how this shapes your temporal lens. Specifically, there are three fundamental aspects of your temporal design that you have to choose, that is: the time frame of your measurements, the time interval between your measurements, and the number of measurement occasions. The specific combination of these three aspects determines the total time span of your study, the degree to which the ongoing process is covered by measurements, and the granularity of your study. Hence, it is important to think about the specific combination you want to use in your study, to ensure that your temporal lens is attuned to the process you want to investigate.
To clarify how the three aspects of the temporal design may vary across different longitudinal studies, the examples in this article concern not only studies based on intensive longitudinal data (ILD), but also on panel data. Furthermore, the designs that are referred to below fall in the categories [interval contingent] (i.e., measurements are taken at regular intervals, such as a daily diary study) and [signal contingent] (i.e., measurements are taken when the researcher gives a signal, as is typical in ESM/EMA or daily diary studies). The information in this article is less relevant for [event contingent] designs, where the measurements are taking place based on encountering certain situations or experiences (e.g., when a participant has to answer questions each time they had a social interaction or when they smoked a cigarette).
In this article, you will find: 1) three key aspects of a temporal design; 2) how these three aspects combine to determine the time span of a study; 3) how the combination of two of them determines the degree to which a process is covered by measurements; and 4) how the combination of all three aspects determines the granularity of your temporal lens.
1 Three key design aspects in longitudinal research
We can distinguish between three key aspects of longitudinal measurement designs (Hamaker, 2023), that is: a) the time frame used in the measurement instrument; b) the time interval between measurements; and c) the number of measurements.
1.1 Time frame of the measurement instrument
A first key aspect of your temporal design is the time frame of your measurement instrument, that is, the specific time period to which an observation pertains. For instance, when using self-report, participants are often asked to answer questions with respect to a particular period, for instance: “How happy were you today?”, “Did something stressful happen in the past hour?”, or “How energetic do you feel right now?” Similarly, when obtaining measurements that are not based on self-report, the researcher also has to decide with respect to what time frame to obtain the measures.
The time frame needs to be chosen such that it does not averaging or smooth over the fluctuations within a person across time that are actually of interest. This means that your time frames should not be too long in comparison to the timescale at which the dynamics of interest operate. For instance, if we want to study emotion regulation, then having a time frame of two weeks is probably (much) too long.
Jean-Philippe wants to study the fluctuations in closeness that people experience towards their partner. He wonders whether to ask participants to answer the question with reference to the current moment (i.e., how close do you feel to your partner right now?), the past hour (i.e., how close did you feel to your partner on average during the past hour?), or the past day (i.e., how close did you feel to your partner on average during the past day?).
Jean Philippe realizes that by choosing a longer time frame, he will obtain less temporal detail; but perhaps, extreme temporal detail is actually not necessary for the fluctuations that are most interesting to him.
Ria can get data from a panel study, in which participants were asked: “On average, how confident were you over the past month?”. The measurements are obtained every six months for a period of three years. Ria also has access to the participants’ average grade each semester.
Ria is interested in the dynamics between grades and confidence, but she realizes that the time frames of the measurements she can obtain are not the same. This can have important consequences for what she may get to see—and what not—of the dynamic relation between these two variables. To avoid this, Ria wants to look into the option of getting measures of academic performance that refer to shorter periods, such that it better matches the time frames of the confidence measurements.
Short time frames may seem desirable, as they allow you to obtain measurements that contain more momentary detail. However, when the time frames are too short, they may contain a lot of momentary variability that could be considered measurement error from the perspective of the actual process you are trying to capture. For instance, if we are interested in an adolescent’s well-being (a concept that is concerned with somewhat more general, longer-lasting experiences rather than momentary experiences), then ESM measures may contain more momentary information than what you are interested in.
Hence, considering the timescale at which a process is thought to operate, is critical in deciding what time frame of the the measurements is appropriate.
1.2 Time interval between measurements
A second key aspect of your temporal design is the time interval between measurements. These can be long (e.g., a year, as is common in panel research), or ultra short (e.g., a millisecond, as when you obtain physiological measurements). Furthermore, they may be fixed (i.e., always the same for everyone, as in an interval contingent design), or (semi-)random (i.e., varying across time and people, as in a signal contingent design). The latter is often used in self-report studies based on ESM/EMA, as a way to ensure participants are caught while living their daily life, rather than anticipating the next beep.
Jean-Philippe wants to see whether having had a conflict with one’s partner is followed by changes in feelings of closeness to one’s partner. He considers whether to use daily diary measures, where at the end of the day he asks participants whether they had a conflict with their partner that day and how close they feel to their partner averaged over the past day. While these measures will be at regular time intervals, Jean-Philippe worries that the dynamics of the process that occurs within the day will be lost due to the temporal resolution of his design.
He therefore considers using an ESM approach with multiple measurements throughout the day during which participants are asked how close they are feeling to their partner right now, and whether they had a conflict with their partner since the previous measurement occasion. While this allows for more temporal detail, the latter question is somewhat awkward, as the conflict could have taken place anywhere between 15 minutes ago (which Jean-Philippe uses as the minimum time between two ESM measurement occasions), and 3 hours ago. This implies that when modeling the relation between these variables, it is somehow giving an average result across these different distances in time, making it somewhat hard to interpret.
Ria is interested in the relation between how confident students feel and their academic performance. Since she only has access to semester averages for the latter, she initially thought that it would be best to obtain measurements of confidence every 6 months to match this.
But confidence is measured with reference to the past month, so Ria is now wondering whether it would be helpful to obtain measurements of confidence every month. However, she is not sure how to analyze data consisting of variables that are measured at different frequencies. She therefore checks MATILDA and reads the article about relating constructs with different timescales.
When deciding on the time interval between the measurements, it is important to think about the timescale at which you believe the process of interest operates. If you want to study a process that takes place within a day, for instance from hour to hour, you have to obtain multiple measurements within a day, and the time interval between the measurements therefore needs to be quite a bit shorter than a day. If your interest is in a developmental process that takes place over a period of months, then the time intervals between the measurements can be longer.
1.3 Number of measurement occasions
A third key aspect of your temporal design is the number of measurement occasions that you will include. If you only obtain a few repeated measures, as is done in panel research, you can only obtain a very rough idea of the underlying processes; for instance, you can determine whether it is increasing of decreasing, but you do not have much information in such data to obtain a richer picture of the temporal patterns. Moreover, if you want to know about individual differences in these patterns, you also need more measurement occasions per person than when you are only interested in the average pattern.
Ying is interested in finding out whether there are individual differences in the degree to which people are characterized by an after-lunch dip. Hence, she has to obtain enough measurements within a day but also across multiple days per person, such that she can reliably estimate individual differences in this pattern.
She believes that about five measurements during the day for five days should be enough. However, she realizes that to really be sure, she should consider the kind of analysis that she will use for these data, and conduct a power analysis to determine whether her design is sufficient to detect the effects she is interested in.
Pia uses a daily diary study to see how fulfilled people feel on days that they work compared to days they did not work. She thus needs to make sure that she includes enough of both working and work-free days per person to be able to reliably estimate the (mean) difference.
When thinking about the number of measurement occasions that are needed, this also really requires you to think of the statistical model or technique that you want to use to analyze the data. For instance, an important question is whether you want to use an analysis technique for N=1 data, or that you want to consider a multilevel modeling approach for N>1 data.
Shanaya wants to investigate individual differences in inertia (autoregression) in momentary negative affect, and whether these are related to neuroticism. Hence, she wants to obtain multiple measurements of momentary negative affect within a day, for multiple days, and she wants to obtain that data from multiple individuals. She needs to determine how many measures per day, for how many days, and for how many people she will need.
There may be some trade-off here. For instance, if Shanaya obtains data from more people, she may need fewer repeated measures, and if she obtains more measures per day, she may need fewer days. But to what extent such trade-offs work out, is probably very much dependent on the underlying process and the particular model she will use in the statistical analysis.
Ideally, you are very specific about the model and the particular model aspect that you want to focus on in your analysis (e.g., a certain parameter, or overal model fit). Then you can do a power analysis, or another simulation study to determine the sample size that is needed, both in terms of number of cases as in terms of number of occasions. The trade-off between these two is actually also a topic of research in itself (Schultzberg & Muthén, 2017).
1.4 Conclusion
The three key aspects of your temporal design should not be considered in isolation. Rather, it is their specific combination that determines how much we zoom in or out (i.e., the time span of our study), how much of the ongoing process is captured by our measurements (i.e., coverage of the process), and how much detail we get to see (i.e., the granularity). These three consequences are elaborated on in the following three sections.
2 The time span of your study
The combination of all three aspects determines the time span that is covered by the study. It thus implies how much you zoom in or out to get a more microscoping or satellite view of the process under study (Klonek et al., 2018).
When you are making use of an in-the-moment time frame (i.e., “how engaged are you feeling right now?”) as is common in ESM, the total time span of your study is simply determined by the time between the first and the last measurement. But when you are using a time frame that is referring to a specific period, this implies that your first measurement is actually concerning a period of time before your measurements started.
When you use fixed time intervals between the measurements, the exact formula for the time span that is covered by your study is given by:
time span = (number of measurements - 1) * time interval + time frameYou can see an example of this in Figure 1: It is base on 3 measurement occasions, each with a time frame of 1 week, and a time interval between measurements of 3 weeks. The resulting time span of the study is 7 weeks. If the intervals between the measurements is increased to 4 weeks instead of 3, the time span of the study would increase to 9 weeks.
Sophie asks participants at a daily basis about their stress during the past day. Hence, in principle, the time frame and the time interval are identical.
She obtains these data for 30 days; hence, the time span of her study is 30 days.
Julio obtained stress measurements 3 times, at an interval of 2 years, and with a time frame of 2 weeks. Hence, the total time span of his study is ((3-1)*(2 years) + 2 weeks = ) 4 years and 2 weeks.
It is important to consider the total time span of your study in comparison to the timescale at which the process you are interested operates. For instance, if you want to investigate whether cognitive functioning fluctuates during the menstrual cycle, having a time span of one or two weeks is not enough; you would want to have data for at least one month, and preferable multiple months.
Similarly, if you want to know how adolescents develop their identity, you have to consider how long it takes before you may expect to see notable and meaningful changes. Moreover, you may also be interested in studying rare events, such as having a relapse after recovering from a depressive disorder or an addiction. It will be challenging to think about the time span that is needed to be reasonably certain that the event occurs, at least for some of the participants in your study.
3 Coverage of the process
The combination of time frame of your measurement instrument (i.e., the period to which an observation pertains), and the time interval between your measurements, can give rise to three different scenarios. These are discussed below
3.1 Time frame and time interval of same length
When the time frame of the measurement instrument and the time interval between measurements are of the same length, this implies that the time frames are spaced back-to-back. In that scenario, all parts of the ongoing process are captured by exactly one measurement frame, because each time frame picks up where the previous one left off. This is for instance the case when we have weekly measures and ask individuals about their experiences during the past week.
Brianna is performing a daily diary study, in which participants are asked at the end of every day questions such as: “On average, how energetic did you feel today?”. Initially, she considers this a case of having time frames being spaced back-to-back.
But when thinking about it more carefully, Brianna realizes that there is probably still a part of the ongoing process that is not captured by these measurements. It is unclear how participants interpret “today”: They are likely to consider this to refer to the period since they woke up. But that means that there may be quite a lot of time between their report the previous evening and their waking up the following day that is not captured by the measurements. Hence, the time frames are not really spaced back-to-back.
Patrick is conducting a panel study with measurements being one year apart, asking people to rate their well-being over the past year. Hence, the time frame of the measurements is identical to the time interval between the measurements, such that for the entire time span of the study, every moment of the process is captured by exactly one measurement frame.
3.2 Time frames being shorter than time intervals
Oftentimes, the time frame is shorter than the time interval between the measurements. This is the case when using the momentary time frame (i.e., ‘’right now’’), but is also quite common in panel research. In the latter case, the time frame is often determined by the instrument that is used and that has been validated with that particular specification (e.g., the past week or past two weeks). If the time frame of the measurements is shorter than the time interval between the measurements, a proportion of the underlying process is not captured by your measurements.
Jack obtained stress measurements during 3 waves, at an interval of 2 years, and with a time frame of 2 weeks. Hence, only 1 week in two years is covered by a measurement frame, which means that less than 1% of the ongoing process is captured by the measurements.
Although this is not an uncommon temporal design, Jack is doubtful that these data can be used to learn about the underlying dynamics of stress. He concludes that these data are mostly appropriate to investigate long-term trends (.e.g, linear trajectories over time), or to predict later health outcomes.
An example of the interplay between time frame and time interval is given in Figure 2. It shows the happiness score of an individual fluctuating over time indicated by the gray line, and measurements based on an instrument that asks about experiences during the past week. The instrument is applied every four weeks, as shown by the light blue shaded frames. This shows that only 25% of the ongoing process is captured by the measurement frames, whereas 75% of the process is not covered by the measurements.
You can also choose a longer time frame, for instance two weeks, while keeping the time interval between measurements the same (i.e., 4 weeks). This results in more of the ongoing process being captured with the measurements, as is shown in Figure 3.
Another way to capture more of the ongoing process with your measurements, is by reducing the time interval between your measurements. If you have a time frame of one week, and combine this with a time interval of one week, the entire process is covered as shown in Figure 4.
3.3 Time frames being longer than time intervals
Another possibility is that the time frame is actually longer than the time interval between measurements, which implies the time frames overlap. In panel research it is not unusual to ask participants to rate themselves “in general”; this may result in participants considering themselves for as long as they can remember at each measurement occasion again. Sometimes, no specific time frame is mentioned, for instance, when the participants is asked “how true is the following statement of you?”. In that case, different participants may interpret the question quite differently.
David performs a panel study in which he measures self-esteem; the measurement waves are six months apart. The participants are asked to indicate how true certain statements are about themselves “in general”.
David ponders how individuals actually interpret such a question. Perhaps, individuals consider their entire life history to the extent that they can remember; but they may also consider a somewhat shorter period, like the past two years. Moreover, different individuals may deal with such a question in a different way, which is not really what David is aiming for. But David is quite certain that most participants will think of a time frame that is longer than six months, which implies measurement frames overlap.
Gunnar did an ESM study, in which he asked participants 10 times per day for 30 days how much self-esteem they have in general. When looking at the data afterwards, Gunnar notices that the self-esteem scores of his participants tend to fluctuate very little over the 30 days. Thus, even though he obtained a lot of repeated measurements per person and the time intervals between the measurements are rather short, the measurements do not provide a very detailed image of the process. The reason for this is that the time frame of the measurements overlap.
Had Gunnar used a different instruction instead (e.g., “how much self-esteem did you experience in the past hour” or “how much self-esteem do you have right now”), the same 300 measurement occasions would provide much more detailed information about the ongoing process.
From a process point of view, unspecific and general measurement frames are problematic; they tend to be ambiguous as individual may interpret them differently, or just too general. In ILD research it is probably better to use more explicitly defined time frames instead, especially if you are interested in the dynamics of a process: What you get to see in terms of the patterns in temporal fluctuations in your data depends to a large extent on the timescale to which your measurements pertain.
4 Granularity of your temporal lens
The granularity of your study is strongly related to the number of repeated measures, but also to the time frame and time interval that is used in your temporal design.
Consider the approach illustrated in Figure 5, based on using a time frame of one week, and obtaining measurements every four weeks. The measurements result in an average for the week that they pertain to. In this case, you lose a lot of detail, not only because a lot of the ongoing process is not captured by measurements, but also because all fluctuations and dynamics within a one week time frame are averaged over.
Just increasing the number of measurements does not change this. If you increase the number of measurements while keeping the time frame and time interval as is, the time span of the study would increase. You would get more information, because you have more repeated measures, but for the currently shown part of the process, the amount of information would not increase.
If you reduce the time interval between measurements while increasing the number of measurements, you may get something as shown in Figure 6. This shows that your measurements cover more of the ongoing process, and that you have more information because you have more measurements. But the information carried by each measurement is—at best (Leertouwer et al., 2022)—still the average of an entire week, and all nuance within a week is lost when using this temporal design.
This shows that just increasing the number of measurements does not necessarily lead to the required granularity for the process you are interested in and the research question you are trying to answer. In thinking about how to adjust your temporal lens such that you can see the detail that is required in your study, you have to consider the number of measurements in combination with the time frame of your measurements and the time interval between your measurements.
These considerations also come into play when thinking of the difference between ESM (based on multiple momentary measures throughout the day), and daily diaries (based on end of day measures that pertain to the entire day). While it may seem at first that the former should be preferred as it can provide more detailed information about the ongoing process, you have to question yourself whether that amount of detail is really needed.
Pia is interested in studying whether people experience more fulfillment on days that they work than on days they do not work. Hence, she performs a daily diary study, in which people report at the end of the day whether or not they worked that day, and how fulfilled they felt that day.
In this case, measuring more often (as would be the case in an ESM/EMA design) would not be beneficial for what Pia is trying to capture. Hence, a daily diary study with a time interval of one day between the measurements is sufficient here.
Furthermore, you should also think about what granularity is needed to obtain the kind of detail that is required to achieve your research goal. For instance, if you are mostly interested in an average pattern across individuals, you need less granularity than if you want to know about individual differences in such patterns, or if you want to study one specific individual’s pattern.
Ying is interested in finding out whether feelings of joy peak in the middle of the day, such that the profile looks like an inverted U-shape. This implies that she should obtain enough measurements across the day; if only two measurements per day were taken, this pattern could not be detected.
Hence, the time interval between the measurements should not be too long, such that within a day, enough measurements are obtained to detect whether this pattern exists. But Ying also needs to consider how many of these cycles she wants to observe: One day may be enough if she is only interested in an average across many individuals, but if she is interested in this pattern for a particular person, she needs to obtain such data for quite a few days.
Hence, if you have a small number of measurement occasions, the granularity of your study will be low. When the number of measurement occasions is large, your study’s granularity may be high, but this is not guaranteed: It also depends on the other design aspects, how these attune to the timescale of the process you want to study, and the research goal that you have in mind.
5 Think more about
When using self-report, the time frame of your measurement instrument requires a mental evaluation from the participant. When the time frame is very short, then it is assumed that people make use of their episodic memory to arrive at an answer; as the time frame increases, it is assumed that individuals at some point switch to their semantic memory (Robinson & Clore, 2002). This has also been related to the difference between the [experiencing self and the remembering self] (Kahneman & Riis, 2004), and is a study topic in itself (cf. Leertouwer et al., 2021).
Moreover, the process that you are interested in is likely to operate at multiple timescales. For instance, when you are interested in affect regulation, you may want to view this through a microscopic lens (Klonek et al., 2018), and obtain measurements that can be used to detect very detailed changes, for instance during a conversation between spouses. But you may also decide that you are more interested in the process of affect regulation as it unfolds across days over the period of a few weeks, and whether specific behaviors such as exercise or alcohol consumption have an effect on it within the same day or the following days. You can also zoom out even further and consider affect regulation across the months or even across the lifespan (Klonek et al., 2018).
You can also decide to investigate how a process manifests at different timescales. Boele et al. (2023) and Bülow et al. (2025) compared the results obtained from datasets and analyses that were based on different time frames of the measurements and different time intervals between the measurements, to evaluate how the results change when these design aspects change.
Finally, practical factors such as participant burden and available funding and personnel will of course also influence your decisions about the feasibility of the ideal temporal design. Such considerations and limitations may require you to adjust accordingly.
6 Takeaway
Your temporal design is to a large extent determined by the time frame of your measurements, the time interval between your measurements, and the number of measurement occasions. Combined, these aspects determine the total time span of your study, the degree to which the ongoing process is covered by the measurements, and the granularity of your study. Hence, you cannot think about these aspects in isolation, but should always consider how their combination helps to attune your temporal lens to what it is you want to learn more about in the process you are studying.
7 Further reading
We have collected various topics for you to read more about below.
- [Retrospective, ecological retrospective, and ecological momentary assessment]
- [Remembering versus experiencing self]
- [Designing a measurement instrument to target between-person or within-person variance]
- [Measurement burst designs]
- [Event contingent bursts]
Acknowledgments
This work was supported by the European Research Council (ERC) Consolidator Grant awarded to E. L. Hamaker (ERC-2019-COG-865468).
References
Citation
@article{hamaker2025,
author = {Hamaker, Ellen L.},
title = {Temporal Design and Its Consequences},
journal = {MATILDA},
number = {2025-05-23},
date = {2025-05-23},
url = {https://matilda.fss.uu.nl/articles/temporal-design-and-its-consequences.html},
langid = {en}
}