This article has Open Peer Review reports available.
Which activity monitor to use? Validity, reproducibility and user friendliness of three activity monitors
© Berendsen et al.; licensee BioMed Central Ltd. 2014
Received: 4 February 2014
Accepted: 15 July 2014
Published: 24 July 2014
Health is associated with amount of daily physical activity. Recently, the identification of sedentary time as an independent factor, has gained interest. A valid and easy to use activity monitor is needed to objectively investigate the relationship between physical activity, sedentary time and health. We compared validity and reproducibility of physical activity measurement and posture identification of three activity monitors, as well as user friendliness.
Healthy volunteers wore three activity monitors simultaneously: ActivPAL3, ActiGraphGT3X and CAM. Data were acquired under both controlled (n = 5) and free-living conditions (n = 9). The controlled laboratory measurement, that included standardized walking intensity and posture allocation, was performed twice. User friendliness was evaluated with a questionnaire. Posture classification was compared with direct observation (controlled measurement) and with diaries (free living). Accelerometer intensity accuracy was tested by correlations with walking speed. User friendliness was compared between activity monitors.
Reproducibility was at least substantial in all monitors. The difference between the two CAM measurements increased with walking intensity. Amount of correct posture classification by ActivPAL3 was 100.0% (kappa 0.98), 33.9% by ActiGraphGT3X (kappa 0.29) and 100.0% by CAM (kappa 0.99). Correlations between accelerometer intensity and walking speed were 0.98 for ActivPAL3, 1.00 for ActiGraphGT3X and 0.98 for CAM. ICCs between activity monitors and diary were 0.98 in ActivPAL3, 0.59 and 0.96 in ActiGraphGT3X and 0.98 in CAM. ActivPAL3 and ActiGraphGT3X had higher user friendliness scores than the CAM.
The ActivPAL3 is valid, reproducible and user friendly. The posture classification by the ActiGraphGT3X is not valid, but reflection of walking intensity and user friendliness are good. The CAM is valid; however, reproducibility at higher walking intensity and user friendliness might cause problems. Further validity studies in free living are recommended.
KeywordsAccelerometer Wearing comfort Posture classification Sedentary Feasibility Reliability Physical activity measurement
Growing evidence shows the negative influence of both physical inactivity and sedentary behavior on health. It has been estimated that physical inactivity is currently related to 6% of mortality and is the main cause of 21-30% of several chronic diseases globally . In addition, an Australian study suggested that 7% of deaths were attributable to prolonged sitting . Recent studies suggest that an increase of physical activity could reduce metabolic risk independent of weight loss or aerobic fitness [3, 4]. In line with this, an increasing amount of evidence reveals an independent association between sedentary behavior and various health outcome measures [2, 5, 6]. However, the optimal amount, frequency and intensity of physical activity and the maximum amount and optimal distribution of sedentary time are still a matter of debate.
Reliable and valid measurements of physical activity and sedentary behavior are essential to draw sound conclusions about their influence on health. However, studies aimed at measuring sedentary behavior have often used self-reported data that suffer from subjectivity [7–9]. Both reproducibility and validity of self-report physical activity and sedentary behavior are variable [9, 10]. Accelerometry has been proposed as a method to objectively quantify sedentary behavior in addition to generally used measures of physical activity [11, 12]. Generally, accelerometers present counts per minute as an intensity outcome based on the accelerations. Previously, the counts per minute output has been tested and used to estimate sedentary time and activity [13, 14]. A problem of this approach is the inability to discriminate between sedentary time and standing time [15, 16]. Recently, several tri-axial activity monitors have been developed that enable measurement of posture (e.g. sedentary behavior and standing) by means of an inclinometer. The ActivPAL3™ (AP; PAL Technologies Ltd, Glasgow, UK), ActiGraphGT3X (AG; ActiGraph LLC, Pensalcola, FL, USA) and CAM (Maastricht Instruments BV, Maastricht, NL) are activity monitors which measure physical activity intensity, register time spent in different postures (e.g. lying, sitting and standing) and thereby assess sedentary time. The AP and the AG have often been used in epidemiological studies, whereas the CAM is a new device developed to provide raw acceleration data. Reproducibility and validity of this inclinometer function has rarely been studied. The posture classification by the CAM was validated in patients with chronic obstructive pulmonary disease and chronic heart failure in daily routine at home . The inclinometer function of the AG showed limited validity and a dependence on location of application (hip vs. back) [13, 18]. Although several validation studies of the inclinometer function of the earlier manufactured uniaxial AP showed good posture classification [14, 15, 19–21], we are not aware of a study aimed at the validity of the posture classification function of the triaxial AP.
The validity and reliability of accelerometry measurements rely on wearing time . However, the required hours per day and total days of measurement are not always met by all participants, which will lead to exclusion of data. Sufficient wearing comfort is a crucial factor in compliance and can consequently affect data quality and validity [23, 24]. Consequently, assessment of wearing comfort and attachment difficulty has been advised .
The aims of this study were to assess 1) reproducibility and validity of walking intensity and the posture classification of the AP, AG and CAM under laboratory conditions; 2) concurrent validity of the AP, AG and CAM with an activity diary in free living and 3) user friendliness of the three activity monitors.
Data were acquired in both controlled and free-living measurements. In the laboratory measurement we compared data with observation, the gold standard; while the free-living measurements provided information in real daily life activities. In the laboratory measurement, the participants were instructed to follow a strict activity and posture protocol in a fixed setting. In the free-living measurement, participants were instructed to write down their activities in a diary every 15 minutes while wearing the devices in daily living. All participants completed a user friendliness questionnaire directly after the laboratory measurement or after returning the activity monitors when participating in the free-living measurement.
A convenience sample of 14 healthy adults with normal BMI participated in the study. Five of them participated in the laboratory measurement (4 male, 1 female, mean age 22.4 years ± 2.2; mean BMI 22.3 ± 1.8); and nine participated in the free-living measurement (4 male, 5 female, mean age 27.2 years ± 8.3; mean BMI 21.3 ± 1.8). Informed consent from participants was obtained. This study was approved by the ethics committee of Maastricht University Medical Centre.
Characteristics of the activity monitors and software
53 × 35 × 7 mm
38 × 37 × 18 mm
63 × 45 × 18 mm
ActivPAL software 6.0.2
Custom Matlab program
Metabolic Equivalent (MET)
Integrated Magnitude Area (IMA)
Yes (inclinometer code)
The AP was taped to the skin at the thigh, using double adhesive PALstickies™ in the laboratory measurement. In the free-living measurement, the AP was waterproofed and attached with a Tegaderm™ dressing (3 M Healthcare, St. Paul, MN, USA); and participants were instructed not to remove it for sleeping or showering. The AG was worn at the waist by means of an elastic belt and the participants were instructed to wear it at their back. As the AG is not waterproof, the device was to be removed when there was a risk of getting wet and during sleeping. To process the AG data, the ActiLife low frequency extension was used. The CAM was worn in an elastic belt around the thigh; also this device was to be removed during sleeping and when there was a risk of getting wet, because it is not waterproof.
The AP and CAM classify time as sitting/lying, standing and activity. The inclinometer function of the AG classifies time as sitting, lying and upright. For the analyses of the activity monitors individually, we assessed all classifications provided. In addition, we used sitting/lying time and upright time as generic measures in the laboratory measurement, to allow comparison between the three activity monitors. Sitting/lying time was defined as lying and sitting postures (regardless whether sitting time was misclassified as lying and vice versa by the AG inclinometer); and upright time was defined as all time spent in an upright orientation (regardless whether active time was misclassified as standing and vice versa by the AP and CAM). Besides the inclinometer function, the AG also discriminates between static posture (lying, sitting and standing) and activity based on a cut point of 100 counts on the vertical axis. For the AG only, the validity of this cut point was assessed in the free-living measurement.
We evaluated four methods in the free living experiment (AP, AG inclinometer, AG counts and CAM). During the free-living measurement, participants wore the three activity monitors simultaneously for at least 3 days. All activity monitors were set to measure 24 hours per day. Participants filled out an activity diary every 15 minutes from waking up till going to bed, writing down the amount of minutes spent in four categories: sitting, walking, standing and other activities. These four categories were then classified as sitting/lying, standing and active. When activities occurred in only one category for longer than 15 minutes, participants were allowed to report them after the subsequent transition. Agreement with the diary was analysed per day. Minutes spent in each category were summed to a total day score. If the amount of minutes per hour registered in the diary exceeded or did not reach 60 minutes, minutes per category were normalised to match 60 minutes in total (referred to as corrected diary data). Both original and corrected diary data were used as comparator for the classification by the activity monitors in free living.
User friendliness questionnaire
The questions within each category of the user friendliness questionnaire
Self-positioning and removal
1. The activity monitor is easy to apply/position
2. The activity monitor is easy to remove
3. The activity monitor is difficult to apply (recoded)
Awareness of wearing
4. The activity monitor fits easily underneath clothing
5. I forgot I was wearing the activity monitor
6. I noticed wearing the activity monitor while doing my daily activities (recoded)
Limitations in behavior
7. The activity monitor limits me during my daily activities (recoded)
8. The activity monitor limits me when I’m exercising (recoded)
9. I’ve changed my activity pattern because of the activity monitor (recoded)
10. I would recommend the activity monitor
11. I would be ashamed if others would see I was wearing the activity monitor (recoded)
The reproducibility of posture classification during the laboratory measurement was analysed on a second-by-second basis with Cohen’s kappa for nominal data, for each activity monitor individually. A kappa-value of < 0.4 was defined as low agreement, > 0.4 was moderate, > 0.6 was substantial and > 0.8 was almost perfect agreement . The reproducibility of the mean intensity of walking during the treadmill exercise was assessed with Intra Class Correlation (ICC) and Bland Altman plots.
Observation was used as gold standard in the laboratory measurement. Data from both laboratory measurements were pooled for validity analyses. Percentages of correctly classified seconds by each activity monitor were calculated and Cohen’s kappa was used to test agreement with the protocol on a second-by-second basis. Friedman’s ANOVA assessed whether the percentages of correctly classified sitting/lying and upright time differed between the three activity monitors. Correlations between walking speed and mean intensity per participant as provided by the standard software were calculated. Concurrent validity between posture classification by the activity monitors in the free-living measurements and the diaries was assessed with ICC and Bland Altman plots. The CAM and AG were only worn during wake time; therefore, their analyses were performed on wake time diary data.
Differences in the category scores of user friendliness between activity monitors were tested with Friedman’s ANOVA and Wilcoxon signed rank test (with an adjusted significance level of p < 0.0167). In addition, compliance in the free living measurement was registered.
Data were described as mean ± SD; if data was not distributed normally, median and 25th and 75th percentile were calculated. Statistical analyses were performed with SPSS version 19 and with a two-tailed significance level of 0.05 (unless mentioned differently).
Correct classification in the laboratory measurement for each category specifically
Classification of sitting and lying time by the ActiGraphGT3X in percentages for each participant (1–5)
Friedman ANOVAs showed that the ability to classify sitting/lying and upright time differed between the three activity monitors (sedentary: p = 0.010; upright time: p = 0.007), in which the AP and CAM performed similarly and the AG had a lower percentage of correct classification in both categories (Table 3).
The validity analyses of the intensity measures resulted in ICCs of respectively 0.98 (CI: 0.97 - 1.00), 1.00 (CI: 1.00 - 1.00) and 0.98 (CI: 0.97 - 1.00) between the treadmill walking speed and mean intensity measures of the AP, AG and CAM (all p < 0.001).
One participant of the laboratory measurement preferred the AP and four participants preferred the AG. Seven participants of the free-living measurements preferred the AP and two preferred the AG. None of the participants indicated CAM as preferred activity monitor to wear.
During the laboratory measurement one participant found it uncomfortable to remove the AP after a short period of measuring and two participants commented that the elastic belt of the CAM was uncomfortable. Following user friendliness issues occurred during the free-living measurements: reported skin irritation due to adhesive material of the AP (n = 3), AG was uncomfortable during sitting, lying or carrying a bag (n =5), skin irritation due to the elastic belt of the CAM (n = 2), aching muscles due to the elastic belt of the CAM (n = 1), CAM was uncomfortable due to sweating while playing sports, not fitting under clothes and did not stay in place (n = 3).
Choosing a suitable activity monitor for scientific studies depends on various aspects. This study aimed to address validity, reproducibility and user friendliness of three activity monitors available for measurement of physical activity and posture classification. Findings of our study indicate a trade-off between these three aspects in the AG and CAM. The AG shows moderate to high reproducibility but low validity for posture allocation and high user friendliness. The CAM shows moderate to high reproducibility, high validity, but low user friendliness. The AP scored well on all three aspects considered: high reproducibility, high validity and high user friendliness (despite reported skin irritation in four participants).
Both AP and CAM showed very good estimations of sitting/lying, standing and walking time. The postures were almost always classified correctly, indicating high validity. Other studies have shown this as well for CAM  and the uni-axial version of the AP [14, 15, 19, 20, 26]. The high reproducibility of the AP was in accordance with findings of a study aimed at the step counts of the uni-axial AP . In the current study, reproducibility of the activity intensity estimated by the CAM at higher walking speed might be insufficient. This raises the question whether the CAM is able to adequately estimate activity intensity at higher intensities, a prerequisite for the discrimination of moderate and vigorous physical activity in pre-post measurements. Bearing in mind that the reproducibility analyses included data of only five participants, the fixation of the CAM by means of the elastic belt might not be secure enough and may have caused the low reproducibility at higher activity intensity.
The ICCs confidence intervals of the AP, AG counts and CAM were acceptable. However, the confidence interval of the ICC of the AG inclinometer function was wide, limiting generalizability to the population level. In addition, plots showed that differences with diary registration were large, despite the moderate to high ICC-values of classification by the activity monitors in daily living. The design of the free living part of the study refrains us from concluding whether the discrepancies were caused by misclassification of the devices or by inaccuracy of the diary as comparator. Participants were asked to report their activities every 15 minutes, as this was believed to be both feasible and accurate. Participants made an effort to report their daily activities in detail (i.e. in minutes precise). Nevertheless, reporting accuracy remains an issue which was not controlled for.
The AG inclinometer did not perform well in terms of reproducibility and validity of posture classification in both the lab and the free-living measurement. The second-by-second analysis of the laboratory measurement showed that much lying time is wrongly classified as non-wear by the inclinometer and sitting and upright time are often mingled. In addition, the amount and type of misclassification seems to be different between participants, for instance, in one participant 83.2% of sitting time was classified correctly, while sitting time in other participants was never classified correctly. The participants were instructed to wear the AG at the back during the measurements in this study because acceleration data reflects physical activity best when the device is worn at the lower back . Although the AG manual states that the inclinometer function performs best when the AG is worn at the hip area, our findings are in line with the results of McMahon and colleagues who evaluated the validity of the inclinometer function when attached at back, waist and upper leg. The results of McMahon and colleagues indicated that compared to the waist, attachment to the back lead to more correctly classified standing time and less correctly classified sitting and lying time. Moreover, neither attachment location led to sufficiently correct sitting and lying identification . Another study in which the AG was worn at the hip found correct posture classifications of 60.6% (standing) to 66.7% (lying). In that study, lying time, watching TV and sitting behind computer were also often classified as non-wear (respectively 14.3%, 6.5% and 9.3%). Also, watching TV and sitting behind computer were often classified as standing time (30.1% and 23.6%) . Most remarkable is the amount of wrongly identified non-wear regardless of attachment location, especially in lying time. In our study, we adopted the non-wear classification provided by the inclinometer function. Usually, non-wear is identified with an algorithm based on a certain amount of inactivity [29–31]. These algorithms have been proven to be sufficiently valid to recognize non-wear in AG measurements . Therefore, it might be advisable to reconsider the added value of the non-wear classification based on inclinometer data. In contrast to the inclinometer function, the discrimination of static and active time based on the cut point of 100 counts on the vertical axis was good. This is in agreement with previous studies [13, 14], which shows that when amount of activity is point of interest, regardless of sedentary time, the AG provides valid data.
Our user friendliness questionnaire addressed five aspects of which three have been proposed earlier. Application of activity monitors in free living requires a device that is easy to use, comfortable and unobtrusive [23, 24]. The CAM scores lowest in most subscales. Possibly, low scores decrease compliance and affect reflection of (in)activity patterns, due to obtrusiveness. The obtrusiveness of the CAM might be higher than the other two devices because of the relatively large size of the CAM and the large elastic belt that was used to wear it. However, compliance of participants to wear the activity monitors in this study was equal. This implies that the application method, removable (CAM and AG) or taped to the skin (AP), does not relate to compliance of wearing. Certain characteristics or subscales of user friendliness might be less or more important dependent on the goal and design of the study. In short measurements, the AG is preferred, whilst the AP is preferred in measurements of several days, even though skin irritation was reported by some participants. Further work is needed to relate the user friendliness to wearing compliance and behavioral adaptations.
The relative small sample size is a limitation of the current study. In addition, the sample consisted of only normal-weight, healthy adults. Therefore, results cannot be generalised to clinical or overweight adults and the user friendliness questionnaire should be assessed for validity and reproducibility in a larger, more variable population. Another limitation is the aforementioned lack of direct observation during free-living measurements. Machado-Rodrigues and colleagues showed that a more detailed diary yielded valid results against an accelerometer . However, diaries always suffer from approximation and although participants were instructed to fill in their diary continuously, we could not control for recall bias in case of non-compliance. Therefore, it is not possible to draw solid conclusions about construct validity from these findings. Nevertheless, by including both controlled laboratory measurements and free-living measurements, our results give an indication of the reproducibility, validity and user friendliness of the three activity monitors.
Results of activity monitoring depend on the device used, and choice of device should depend on the research aims and design. The majority of the studies which led to the current consensus on the negative influence of sedentary time on health, independent of physical activity, are based on subjective measures. As an objective measure, accelerometry can reinforce earlier results. The current study shows that the AP and CAM are able to classify posture and that the inclinometer function of the AG provides no valid posture classification. However, the AG can well be used if level of physical activity is of interest.
The study is part of a project funded by ZonMW, The Netherlands Organization for Health Research and Development (project number: 123000002).
- World Health Organization: Global Health Risks: Mortality and Burden of Disease Attributable to Selected Major Risks. 2009, Geneva, Switzerland: WHO Press, 10-11.Google Scholar
- van der Ploeg HP, Chey T, Korda RJ, Banks E, Bauman A: Sitting time and all-cause mortality risk in 222 497 Australian adults. Arch Intern Med. 2012, 172: 494-500.View ArticlePubMedGoogle Scholar
- Duvivier BM, Schaper NC, Bremers MA, van Crombrugge G, Menheere PP, Kars M, Savelberg HH: Minimal intensity physical activity (standing and walking) of longer duration improves insulin action and plasma lipids more than shorter periods of moderate to vigorous exercise (cycling) in sedentary subjects when energy expenditure is comparable. PLoS One. 2013, 8: e55542-View ArticlePubMedPubMed CentralGoogle Scholar
- Ekelund U, Franks PW, Sharp S, Brage S, Wareham NJ: Increase in physical activity energy expenditure is associated with reduced metabolic risk independent of change in fatness and fitness. Diabetes Care. 2007, 30: 2101-2106.View ArticlePubMedGoogle Scholar
- Bankoski A, Harris TB, McClain JJ, Brychta RJ, Caserotti P, Chen KY, Berrigan D, Troiano RP, Koster A: Sedentary activity associated with metabolic syndrome independent of physical activity. Diabetes Care. 2011, 34: 497-503.View ArticlePubMedPubMed CentralGoogle Scholar
- Hamilton MT, Hamilton DG, Zderic TW: Role of low energy expenditure and sitting in obesity, metabolic syndrome, type 2 diabetes, and cardiovascular disease. Diabetes. 2007, 56: 2655-2667.View ArticlePubMedGoogle Scholar
- Healy GN, Clark BK, Winkler EA, Gardiner PA, Brown WJ, Matthews CE: Measurement of adults’ sedentary time in population-based studies. Am J Prev Med. 2011, 41: 216-227.View ArticlePubMedPubMed CentralGoogle Scholar
- Thorp AA, Owen N, Neuhaus M, Dunstan DW: Sedentary behaviors and subsequent health outcomes in adults a systematic review of longitudinal studies, 1996–2011. Am J Prev Med. 2011, 41: 207-215.View ArticlePubMedGoogle Scholar
- Clark BK, Sugiyama T, Healy GN, Salmon J, Dunstan DW, Owen N: Validity and reliability of measures of television viewing time and other non-occupational sedentary behaviour of adults: a review. Obes Rev. 2009, 10: 7-16.View ArticlePubMedGoogle Scholar
- Prince SA, Adamo KB, Hamel ME, Hardt J, Connor Gorber S, Tremblay M: A comparison of direct versus self-report measures for assessing physical activity in adults: a systematic review. Int J Behav Nutr Phys Act. 2008, 5: 56-View ArticlePubMedPubMed CentralGoogle Scholar
- Foerster F, Fahrenberg J: Motion pattern and posture: correctly assessed by calibrated accelerometers. Behav Res Methods Instrum Comput. 2000, 32: 450-457.View ArticlePubMedGoogle Scholar
- Oliver M, Schofield GM, Badland HM, Shepherd J: Utility of accelerometer thresholds for classifying sitting in office workers. Prev Med. 2010, 51: 357-360.View ArticlePubMedGoogle Scholar
- Carr LJ, Mahar MT: Accuracy of intensity and inclinometer output of three activity monitors for identification of sedentary behavior and light-intensity activity. J Obes 2012. 2012, 2012: 460271-Google Scholar
- Kozey-Keadle S, Libertine A, Lyden K, Staudenmayer J, Freedson PS: Validation of wearable monitors for assessing sedentary behavior. Med Sci Sports Exerc. 2011, 43: 1561-1567.View ArticlePubMedGoogle Scholar
- Kozey-Keadle S, Libertine A, Staudenmayer J, Freedson P: The Feasibility of Reducing and Measuring Sedentary Time among Overweight, Non-Exercising Office Workers. J Obes. 2012, 2012: 282303-View ArticlePubMedGoogle Scholar
- Freedson P, Bowles HR, Troiano R, Haskell W: Assessment of physical activity using wearable monitors: recommendations for monitor calibration and use in the field. Med Sci Sports Exerc. 2012, 44: S1-4.View ArticlePubMedPubMed CentralGoogle Scholar
- Annegarn J, Spruit MA, Uszko-Lencer NH, Vanbelle S, Savelberg HH, Schols AM, Wouters EF, Meijer K: Objective physical activity assessment in patients with chronic organ failure: a validation study of a new single-unit activity monitor. Arch Phys Med Rehabil. 2011, 92: 1852-1857. e1851View ArticlePubMedGoogle Scholar
- McMahon GC, Brychta RJ, Chen KY: Validation of the Actigraph (GT3X) inclinometer function. Med Sci Sports Exerc. 2010, 42: 489-View ArticleGoogle Scholar
- Grant PM, Ryan CG, Tigbe WW, Granat MH: The validation of a novel activity monitor in the measurement of posture and motion during everyday activities. Br J Sports Med. 2006, 40: 992-997.View ArticlePubMedPubMed CentralGoogle Scholar
- Hart TL, Ainsworth BE, Tudor-Locke C: Objective and subjective measures of sedentary behavior and physical activity. Med Sci Sports Exerc. 2011, 43: 449-456.View ArticlePubMedGoogle Scholar
- Godfrey A, Culhane KM, Lyons GM: Comparison of the performance of the activPAL Professional physical activity logger to a discrete accelerometer-based activity monitor. Med Eng Phys. 2007, 29: 930-934.View ArticlePubMedGoogle Scholar
- Corder K, Brage S, Ekelund U: Accelerometers and pedometers: methodology and clinical application. Curr Opin Clin Nutr Metab Care. 2007, 10: 597-603.View ArticlePubMedGoogle Scholar
- Mathie MJ, Coster AC, Lovell NH, Celler BG: Accelerometry: providing an integrated, practical method for long-term, ambulatory monitoring of human movement. Physiol Meas. 2004, 25: R1-20.View ArticlePubMedGoogle Scholar
- Trost SG, McIver KL, Pate RR: Conducting accelerometer-based activity assessments in field-based research. Med Sci Sports Exerc. 2005, 37: S531-543.View ArticlePubMedGoogle Scholar
- Landis JR, Koch GG: The measurement of observer agreement for categorical data. Biometrics. 1977, 33: 159-174.View ArticlePubMedGoogle Scholar
- Maddocks M, Petrou A, Skipper L, Wilcock A: Validity of three accelerometers during treadmill walking and motor vehicle travel. Br J Sports Med. 2010, 44: 606-608.View ArticlePubMedGoogle Scholar
- Dahlgren G, Carlsson D, Moorhead A, Hager-Ross C, McDonough SM: Test-retest reliability of step counts with the ActivPAL device in common daily activities. Gait Posture. 2010, 32: 386-390.View ArticlePubMedGoogle Scholar
- Bouten CV, Sauren AA, Verduin M, Janssen JD: Effects of placement and orientation of body-fixed accelerometers on the assessment of energy expenditure during walking. Med Biol Eng Comput. 1997, 35: 50-56.View ArticlePubMedGoogle Scholar
- Troiano RP, Berrigan D, Dodd KW, Masse LC, Tilert T, McDowell M: Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc. 2008, 40: 181-188.View ArticlePubMedGoogle Scholar
- Winkler EA, Gardiner PA, Clark BK, Matthews CE, Owen N, Healy GN: Identifying sedentary time using automated estimates of accelerometer wear time. Br J Sports Med. 2012, 46: 436-442.View ArticlePubMedGoogle Scholar
- Choi L, Ward SC, Schnelle JF, Buchowski MS: Assessment of wear/nonwear time classification algorithms for triaxial accelerometer. Med Sci Sports Exerc. 2012, 44: 2009-2016.View ArticlePubMedPubMed CentralGoogle Scholar
- Machado-Rodrigues AM, Figueiredo AJ, Mota J, Cumming SP, Eisenmann JC, Malina RM, Coelho ESMJ: Concurrent validation of estimated activity energy expenditure using a 3-day diary and accelerometry in adolescents. Scand J Med Sci Sports. 2012, 22: 259-264.View ArticlePubMedGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2458/14/749/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.