• Introduction
• Habitual physical activity
• Diurnal variation of activity
• Between-day variability
• Seasonal variation
• Pattern metrics and methods
• Lifetime variation in activity
• References

Sources of variance in physical activity include inter-individual or between-individual variation, which reflects the differences in behaviour between individuals, and intra-individual or within-individual variation, which reflects the differences in behaviour within each person. For direct measurement of variability, it is most often the within-person variability that is being assessed since most measurement instruments are applied to individuals, and the between-individual variability can then be quantified in the analysis of datasets containing multiple individuals; however for measurement methods where the object of measurement is not one person but a group of individuals (street imagery, observation of school classes, etc), variability between individuals could be directly assessed at the point of measurement.  

Variability in behaviour within a person is often defined with respect to time but it can also be defined as for example diversity of behaviour types, ie engaging in both aerobic and muscle-strength enhancing activities as expressed in the physical activity guidelines, or taking part in both recreational exercise purely for fun and competitive sports, individual vs team sports, etc. Collectively, these differences within each person are referred to as the within-individual variation.

With respect to time, physical activity of an individual displays distinct variability over different time-scales; it typically changes from minute to minute, hour to hour, and from one day to another during the week. Patterns may also be different over longer time-scales such as from season to season or from one year to another.  

On the shorter time-scales, they are governed by the time of day and factors such as availability of daylight, work/school patterns and religious practices. On longer time-scales seasonal variation may arise from preference for certain exercise types in winter and summer, timing of different holidays or as part of attending to different subsistence tasks (e.g. agricultural) or construction work. The longest time-scales would include variability across life-stages, e.g. starting and finishing schooling and higher education or apprenticeship training, changing jobs, and retirement. Variability in physical activity is often considered in terms of activity volume but the behaviour can vary by all its sub-dimensions such as intensity distribution, type or context.

Within-individual variation in physical activity with respect to time is often considered an assessment challenge because population health science research is generally more interested in the habitual level of activity rather than the physical activity occurring in any particular one hour or on any single day. However, habitual physical behaviour is a latent variable, as it is often not directly observable. In order to estimate habitual level of activity, we employ sampling designs that capture one or more of the time variance components outlined below, typically capturing at least the shorter time-scale variation components in single measurements and employing repeated measurement designs for capturing the longer time-scale measurement components.

The objective of estimating habitual level of activity is to get to a stable average estimate of the physical behaviour being studied for each person, often attempting to eliminate undue influence of within-person variability. However, the pattern of within-person variation can also be studied in its own right and there are dedicated sections on this page focusing on how we may quantify behavioural variability at specific time-scales.

Monitoring period(s)

If an individual participated in exactly the same physical activity every day, then one day of monitoring would be sufficient to observe the habitual level. If the individual only had two different types of day, then two days of monitoring would be the minimum required to capture the habitual level using the daily average (Baranowski & de Moor, 2000). However, human behaviour is more complex and no two days are truly the same. The greater the difference between days (within-individual variation), the longer duration of monitoring is required to reliably assess the habitual level of activity.

Some questionnaire methods try to account for within-individual variation by assessing physical activity over the past year or longer. A year of monitoring using diary/log methods or body-worn objective methods would likely impose too great a burden. Furthermore, the monitoring period duration for objective methods follows the rule of diminishing returns, so whilst longer monitoring periods are always desirable from an information perspective, the incremental information becomes smaller and smaller. Determining the minimum monitoring period(s) required to reliably capture the latent habitual physical activity level is therefore an important consideration.

Day-to-day variation tends to be greater in children and 4-9 days of monitoring are common. In adults monitoring periods tend to be shorter, e.g. 3-5 days; however 7 days of monitoring may appear more attractive from a face validity point of view if the population under study follows the cultural practices that have a natural period of a week (Doherty et al., 2017; Trost et al., 2005). Different behavioural subcomponents may show different levels of day-to-day variation, which needs consideration when designing measurement protocols aiming to estimate certain subcomponents of interest in a reliable manner.

Repeat assessment

If there is likely to be substantial variability in activity levels between seasons or even longer time-scales, repeat assessments of activity can be built in to the study design at the individual level for example by measuring each person two or more times a year. Alternatively, the impact of these variability components can be taken into account at the point of statistical analysis, a practice often refered to as measurement error correction. This still requires some repeat assessment but only in a smaller sample of the population and it does not need to be the same individuals as in the study where the correction is applied. In essence, such repeat measurement studies quantify the degree to which the behaviour is stable over time, and so are sometimes also refered to as test-retest reliability studies. 

The variation in physical activity over the 24-hour period is one of the most prominent features of physical behaviour, with long blocks of very low activity during nocturnal sleep interspersed with sporadic patterns of activity during the day (Figure P.1.11) . The within-person variability of behaviour at this time-scale is known as the diurnal rhythm and is closely related to the biological concept of the circadian rhythm which denotes the intrinsic cyclical pattern without the influence of external time clues (zeitgebers).

Diurnal variation in behaviour can be captured with simple methods by collecting data for specific periods of the day, e.g., mornings and evenings; however, with the increased use of device measures, continuous variables covering the whole 24-hour period is easily captured. Estimates of diurnal variation include maximum and minimum values within a day, as well as the timing of these. More sophisticated methods can also be applied, for example cosinor analysis, structured functional principal component analysis and behaviour "barcode" methods. 

Knowledge of the diurnal variations in physical activity could provide useful information for planning and implementing interventions aiming at increasing physical activity levels. 

Figure P.1.11 Time-series of physical activity intensity

Physical behaviours also vary between days within the individual. This is acknowledged in the public health recommendations for physical activity which contain behavioural targets that are less frequent than than every day over the course of the week. Figure P.1.10. shows about 5 days of objectively measured physical activity, and whilst there are several similarities between each of the 5 days, there are also some differences, both in terms of volume and number of intensity peaks and their duration. The assessment of between-day variability within a person seeks to quantify such differences. Examples of variables that capture between-day variability include:

• Weekdays versus weekend days
• Weekday versus weekdays
• Weekend day versus other weekend day
• Previous day versus present day
• Periodicity of behaviour (Fourier transformations)

Analysis methods include simple stratification of the data within each person to derive for example the average activity energy expenditure for each day or for weekdays and weekends. Other types of analysis quantify the degree of regularity or periodicity of the behaviour being studied, such as cosinor analysis and behaviour "barcode" methods. Within-day and between-day variability can also be studied together in hierarchical models, indexing time of day at one level and day of week (or month) on another level.    

Most studies relying on device-based measures employ a week-long monitoring protocol for the measurement of physical activity, which provide reliable estimates for that week of the year but provide no information on how behaviour may vary across seasons.

Seasonal patterns are cyclical, with a period of 365 days (or 366 in leap years) and likely contribute to total variability over time. There are several variance components that characterise time of year, one being the availability of daylight which can be described by a single sine function for each latitude (periodicity would be the same for other latitudes but the amplitude would be different). Other characteristics include temperature, precipitation, and wind patterns which are also cyclical but which have additional variance components that are more complex. Several human behaviours align with weather patterns, either by preference or by necessity. Figure P.1.12 shows the seasonal variation in MVPA patterns in populations living in Cameroon and the UK.

Figure P.1.12 Seasonal variation in MVPA in Cameroon and UK. Source: Brage et al., 2020.

Lifetime exposure to physical activity and its variation is an important construct in investigating its association with many chronic diseases with long induction or latency period. Although there are a few instruments that are designed to directly capture physical behaviours across the lifespan up to the point of assessment, the most common way to assess variation in these behaviours is repeat assessment in longitudinal studies.

Collectively, these measurements allow the quantification of change in physical activity from which stronger conclusions can be drawn on the potential causal role of physical behaviours and health outcomes.

1. Atkin AJ, Sharp SJ, Harrison F, Brage S, Van Sluijs EM. Seasonal Variation in Children's Physical Activity and Sedentary Time. Medicine and Science in Sports and Exercise. 2015;48:449-56
2. Baranowski T, de Moor C. How Many Days Was That? Intra-individual Variability and Physical Activity Assessment. Research Quarterly for Exercise and Sport. 2015;71 Suppl 2:74-8
3. Besson H, Harwood CA, Ekelund U, Finucane FM, McDermott CJ, Shaw PJ, Wareham NJ. Validation of the historical adulthood physical activity questionnaire (HAPAQ) against objective measurements of physical activity. The International Journal of Behavioral Nutrition and Physical Activity. 2010;7:54
4. Brage S, Assah F, Msyamboza KP. Quantifying population levels of physical activity in Africa using wearable sensors: implications for global physical activity surveillance. BMJ Open Sport & Exercise Medicine. 2020;6:e000941
5. Chasan-Taber L, Erickson JB, McBride JW, Nasca PC, Chasan-Taber S, Freedson PS. Reproducibility of a self-administered lifetime physical activity questionnaire among female college alumnae. American Journal of Epidemiology. 2002;155:282-9
6. De Vera MA, Ratzlaff C, Doerfling P, Kopec J. Reliability and validity of an internet-based questionnaire measuring lifetime physical activity. American Journal of Epidemiology. 2010;172:1190-8
7. Doherty A, Jackson D, Hammerla N, Plötz T, Olivier P, Granat MH, White T, van Hees VT, Trenell MI, Owen CG, et al. Large Scale Population Assessment of Physical Activity Using Wrist Worn Accelerometers: The UK Biobank Study. PLOS ONE. 2016;12:e0169649
8. Evenson KR, Herring AH, Wen F. Accelerometry-Assessed Latent Class Patterns of Physical Activity and Sedentary Behavior With Mortality. American Journal of Preventive Medicine. 2016;52:135-143
9. Friedenreich CM, Courneya KS, Bryant HE. The lifetime total physical activity questionnaire: development and reliability. Medicine and Science in Sports and Exercise. 1998;30:266-74
10. Matthews CE, Moore SC, George SM, Sampson J, Bowles HR. Improving self-reports of active and sedentary behaviors in large epidemiologic studies. Exercise and Sport Sciences Reviews. 2012;40:118-26
11. Paraschiv-Ionescu A, Perruchoud C, Buchser E, Aminian K. Barcoding human physical activity to assess chronic pain conditions. PLOS ONE. 2011;7:e32239
12. Prioreschi A, Brage S, Westgate K, Micklesfield LK. Describing the diurnal relationships between objectively measured mother and infant physical activity. The International Journal of Behavioral Nutrition and Physical Activity. 2017;15:59
13. Rowlands AV, Gomersall SR, Tudor-Locke C, Bassett DR, Kang M, Fraysse F, Ainsworth B, Olds TS. Introducing novel approaches for examining the variability of individuals' physical activity. Journal of Sports Sciences. 2014;33:457-66
14. Saint-Maurice PF, Sampson JN, Keadle SK, Willis EA, Troiano RP, Matthews CE. Reproducibility of Accelerometer and Posture-derived Measures of Physical Activity. Medicine and Science in Sports and Exercise. 2019;52:876-883
15. Shou H, Zipunnikov V, Crainiceanu CM, Greven S. Structured functional principal component analysis. Biometrics. 2013;71:247-257
16. Trost SG, McIver KL, Pate RR. Conducting accelerometer-based activity assessments in field-based research. Medicine and Science in Sports and Exercise. 2005;37:S531-43
17. Yerrakalva D, Wijndaele K, Hajna S, Westgate K, Khaw KT, Wareham N, Griffin SJ, Brage S. Do older English adults exhibit day-to-day compensation in sedentary time and in prolonged sedentary bouts? An EPIC-Norfolk cohort analysis. PLOS ONE. 2019;14:e0224225