Journal Logo


Identifying Walking Trips Using GPS Data


Author Information
Medicine & Science in Sports & Exercise: February 2011 - Volume 43 - Issue 2 - p 365-372
doi: 10.1249/MSS.0b013e3181ebec3c


Outdoor behavioral patterns have long been an important subject among researchers in the fields of exercise science, transportation, parks and recreation, and planning. Walking is recognized not only as a travel mode but also as an important leisure time physical activity. Moreover, walking has been emphasized as a behavior that improves health outcomes (10,11).

As research on walking has increased, more studies have been conducted to examine the quality of the walking data collected. Accurate measurement of walking allows dose-response relationships between walking and health variables to be identified more easily and precisely (2). The predominant tools for collecting information on walking include direct observation at specific locations and self-report, such as those collected through surveys and diaries (24), which inevitably include reporting errors (2). Other objective instruments, including pedometers and accelerometers, can provide more accurate information about the intensity and duration of walking (7).

Portable global positioning system (GPS) technology provides an innovative means to measure walking. In addition to the measurement of walking behavior, GPS units have a great advantage in contextualizing where physical activity takes place (21,22), which cannot be provided by other objective devices, such as accelerometers or pedometers. This unique characteristic of GPS units can yield a greater understanding of how different physical environment attributes are associated with individuals' behaviors (27).

Benefiting from the matured GPS technology (9,11,22), several studies have tried to classify the type of activity using only GPS data or the combination of GPS and accelerometer data (4,16,23,25,27). However, the processes or algorithms that these studies have developed have not provided a satisfactory solution to characterize walking in free-living conditions because most have been focused on characterizing vehicle-based trips (4,17,18,22,28,29), have tested the algorithm with predefined pedestrian routes or activity schedules (16,23,27), or have been too complicated, resulting in data manipulation and algorithms that required considerable time and proficiency with geographic information systems (GIS) (4,25). The main purpose of this study was to develop an algorithm to identify outdoor walking trips in free-living conditions using only GPS data from a portable GPS unit.



The study proceeded in two phases. The first phase determined and tested criteria that could be used to identify outdoor walking trips from the location data recorded by the GPS units. In 2005, this was done by collecting GPS and self-reported location diary data on a sample of five healthy research staff aged 33-50 yr during 7 d. Three were men and two were women. The resulting 35 person-days of data were used to calibrate the algorithm and identify the criteria that best matched the self-reported walking trips.

Once the best algorithm was identified in the calibration phase, we applied the algorithm to a second independent sample. Forty adult volunteers capable of walking unassisted for 20 min or more were recruited during 2005 to participate in the validation phase of this study. Participants were asked to complete two types of diaries during 4 d (2 d each) and concurrently wear the GPS units except when swimming and bathing. They were instructed to recharge the GPS units at night. A $20 incentive was given to participants who successfully completed 4 d of the study. Four participants did not adhere to the GPS protocol, and two participants withdrew before completing the study and therefore were excluded from the final analysis. Consequently, we had 34 participants' 136 person-days of data. Twenty-one participants were women and 13 were men. Each participant read and signed a participant consent form, and the study was approved by the institutional review board of the Office of Human Research Ethics at University of North Carolina, Chapel Hill.

GPS unit.

The portable GPS unit used in this study was the Garmin Foretrex 201 (Olathe, KS), weighing 78 g and measuring 83.8 × 43.2 × 15.2 mm3. The units are designed to be worn on the wrist, upper arm, or around the waist attached to a belt. Internal nonvolatile memory enabled the unit to store 10,000 location points before the data required downloading. A study using the Garmin Foretrex 201 GPS unit reported that average bias among units ranged from 0.22 to 1.86 m in static conditions and from 10.7 to 20.1 m in free-living conditions (22). The same study concluded that the portable GPS units tested was able to precisely track participants' movements, although built environment characteristics affect the positional accuracy in free-living conditions. Wrist and hip placement of the GPS unit produced similar, high-quality locational data. The units were set to record the positional coordinates of their location at 60-s intervals with the Wide Area Augmentation System enabled to record a maximum 96 h of records for the four study days. The Wide Area Augmentation System consists of satellites and ground stations that correct GPS signal errors caused by atmospheric disturbances, timing, and satellite orbit errors, resulting in increased positional accuracy. Each row of GPS data included date, time, latitude and longitude in degrees (°) and minutes (′), and average speed. We removed data headers, removed data when the unit was unable to record position, converted coordinate information into decimal degrees, and transformed the data into wide-character ASCII format. The GPS units and data manipulation procedures used in the calibration and validation phases were identical.


The diary used for calibration was a location diary that we developed. The purpose of the location diary was to help determine how, when, and why participants travel to different locations. It required that participants fill in the trip start and arrival time, mode of travel, and location of activity in close-ended format (SDC1, Figure. Sample sheet of the location diary; Participants were asked to record any travel, regardless of travel mode, that traversed a distance greater than 91 m (300 ft). Similar to the protocols of National Household Transportation Survey (NHTS) diary (20), walking trips may include jogging activity.

In the validation phase, participants were given one of two travel diaries for the first 2 d: either the location diary, previously described, or the NHTS diary. After 2 d, participants returned the first diary and received the second diary to be filled out while carrying the GPS unit for the following 2 d. Which diary was given first to each participant was determined at random, with half of the participants receiving the NHTS diary first and the other half receiving the location diary second. The NHTS diary required that users fill in the trip destination, trip start and arrival time, purpose of trip, mode of travel, and travel distance as open-ended questions (SDC2, Figure. Sample sheet of 2001 National Household Transportation Survey diary; By using two types of diaries, we intended to examine whether diary type may affect agreement between GPS-derived and diary-reported walking trips.

Calibration of the algorithm.

Identifying walking trips from the GPS data was performed with a two-step process. First was the identification of a trip, regardless of the mode used (motorized or nonmotorized). This was indicated by a group of GPS points that contained a beginning (origin) and ending (destination) set of points. Thus, the maximum time gap between points (criterion 1) was used as the criterion for the first step. Second is the classification of each trip as a walking trip, which resulted in four criteria used: (criterion 2) minimum number of consecutive track-points, (criterion 3) specific speed range, (criterion 4) minimum distance between start and end points of a trip, and (criterion 5) maximum time gaps between trips to be merged (Fig. 1), detailed next.

Flow diagram demonstrating the algorithm to identify outdoor walking trips with corresponding criteria.
  1. GPS units detect satellite signals when they are located outdoors in direct sight of satellite and regularly fail to record positions indoors. Therefore, the absence of location data lasting more than a threshold amount of time indicated that the participants were unlikely to be involved in outdoor walking. We used 3 and 5 min as indicators of when a travel episode ended and a new one began.
  2. The minimum number of consecutive track points indicated the minimum duration of trips. We classified points as an outdoor walking trip when their duration was at least 3 or 5 min. We ignored GPS data that were not consecutively recorded for at least either 3 or 5 min.
  3. Comfortable walking speeds of adults aged 20-79 yr range from 4.5 to 5.3 km·h−1, and maximum speeds, sometimes called exercise walking or health walking speed, range from 6.3 to 9.2 km·h−1. (3). We defined the speed of walking trips using two alternative ranges: 3 km·h−1 < speed < 6 km·h−1 or 2 km·h−1 < speed < 8 km·h−1. After this step, consecutive track points were labeled as preliminary walking trips.
  4. We defined a minimum distance of a trip to ensure that participants were involved in outdoor walking trips instead of standing, but not walking, outdoors. The minimum distance between the start and end points of trip was 30 m (100 ft).
  5. Signal loss for the GPS units could result in data gaps within a single trip. When the next set of points was recorded within 3 or 5 min of the end of the previous set of consecutive points, we presumed that they were part of the same trip separated by signal loss and merged them into a single trip. A gap of more than 3 or 5 min was interpreted as a separate trip.

Table 1 shows the values for the five criteria we tested. Combinations of the five criteria yielded 16 possible algorithms. We applied these 16 algorithms to the 35 person-days of diary and GPS data from the calibration sample.

Summary criteria for defining an outdoor walking trip.

Data analysis.

The best algorithm was the one that had the highest Spearman's (ρ) correlation coefficient for the duration and the number of trips per person per day, as reported in the location diary. We also considered the kappa (κ) statistics to compare the agreement between diary-reported trips and GPS-identified trips.

Following the interpretation of Landis and Koch (15), agreement was considered to be slight when the agreement statistics were below 0.2, fair when agreement was between 0.2 and 0.4, moderate when agreement was between 0.4 and 0.6, substantial when agreement was between 0.6 and 0.8, and almost perfect when agreement was greater than 0.8. STATA SE 9.0 was used for all statistical analysis (13,15).

The best-fitting algorithm as determined in the calibration phase was applied to the validation sample and used to calculate the number of walking trips from the GPS data of this sample. Based on ρ and κ, we compared agreement for the number and duration of walking trips per person per day between walking trip data that were GPS-derived and self-reported. Furthermore, Kolmogorov-Smirnov's test and its D statistics were used to determine whether the distribution of the two data sets differed significantly. The null hypothesis was that the two algorithms result in similar distributions for the outcomes analyzed. A statistically significant D statistics means that the null hypothesis that the distributions are similar can be rejected and that the largest difference between the distribution functions is given by D (6).

Although there may be high agreement in the total number and duration of outdoor walking trips per day identified through the GPS data and the diaries, disagreements may remain regarding specific trips. For example, a walking trip identified from the GPS data may not be reported in a diary. To address this, each individual trip reported in the diary was matched to the GPS data, and each trip in the GPS data was matched to a diary-reported trip. A trip identified from the GPS data was considered to be matched when more than 70% of its duration was found within the time of a reported trip in the diary. A reported trip in the diary was considered to be matched when one or more matched trips identified from the GPS data using the above criterion were found within its reported duration. Percent of matched trips was used to assess the degree of agreement for specific trips. In the validation phase, all trips that were not matched were examined qualitatively one by one, identifying possible explanations for the disagreement using ancillary information. For trips not reported in the diary, we used a GIS (ArcGIS 9.2; Environmental Systems Research Institute, Redlands, CA) to examine the locations of the GPS data. For trips not identified through GPS, we used the diary time and travel mode information to examine the GPS data to infer potential problems in the algorithm.


Calibration phase.

Aggregating across participants, 42 walking trips were reported on the diary. Of these, 35 lasted more than 3 min and 31 lasted more than 5 min. The mean ± SD number of daily reported trips per participants was 1.00 ± 1.46 and 0.89 ± 1.30 for trips lasting more than 3 and 5 min, respectively, and their mean ± SD duration was 18.7 ± 16.6 and 20.3 ± 17.0 min, respectively. The 16 algorithms were identified between 25 and 50 trips (Table 2). The algorithms resulting in the highest number of trips were the ones with a small number of consecutive points required (criterion 2) and a large range of speeds allowable (criterion 3).

Agreement number of walking trips per day and the duration per person per day between GPS-derived data and the self-reported location diary (n = 35 person-days, calibration sample).

Spearman's correlation between number of daily walking trips per person estimated from the algorithms and those reported by the location diary ranged from 0.71 to 0.84. The κ statistics between the two showed a moderate level of agreement, ranging from 0.42 to 0.54. Regarding walking duration per day, Spearman's correlations ranged from 0.83 to 0.93, with an average of 0.88. Among the 16 algorithms, alternatives 11 and 12 showed the highest correlations in both the number and duration of walking trips per day. Kolmogorov-Smirnov's test indicated that alternatives 3, 4, 11, and 12 had the smallest difference between the distributions of diary-reported walking trips per person per day and the GPS-derived walking trips. The same four alternatives had the closest fit between the distributions of duration of walking trips per day (D = 0.14, P = 0.87). A comparison across different algorithms suggested that the fifth criterion, maximum time gap between trips to be merged, had a small effect on the number of trips identified, so the number of walking trips identified by alternatives 11 and 12 was identical. Using the criteria of alternative 11, 26 (90%) of 29 trips identified from the GPS data were found in the diary, whereas 25 (81%) of 31 trips reported in the diary were found in the GPS data. The numbers of matched trips in both cases were not identical because multiple trips could be identified from the GPS data while they were a single self-reported trip.

On the basis of the findings of the calibration data set in Table 2, the preferred algorithm was based on the following characteristics: 1) a maximum time gap between points of 3 min, 2) at least 5 min or more of duration, 3) a speed range from 2 to 8 km·h−1, 4) at least 30 m of displacement between the start and end points of a trip, and 5) a maximum time gap between trips to be merged of 3 min.

Validation phase.

Thirty-four participants reported 178 walking trips in the location and NHTS diaries. A considerable number of those trips were excluded in testing our algorithm because they did not last at least 5 min or contain consecutive GPS records. The final number of self-reported walking trips was 61.

Compared with the Spearman's correlation coefficients in the calibration phase, correlations in the validation phase were slightly lower for the NHTS and location diaries (Table 3). For the number of trips, Spearman's correlation coefficients comparing the NHTS and location diaries were 0.79 and 0.88, respectively. For the duration of trips, the coefficient for the location diary to the GPS (0.89) was higher than the coefficient for the NHTS diary (0.77), but the differences in the number and duration of trips were not statistically significant (P = 0.78 and P = 0.17, respectively). Kolmogorov-Smirnov's test also indicated that the distribution of walking trips per day (D = 0.04, P = 1.00) and duration per day (D = 0.15, P = 0.08) was not different between GPS and the diaries. Compared with the results in the calibration phase, the κ statistics in the validation phase was generally higher.

Agreement (κ and ρ with 95% CI) of number of walking trips per day and duration per person per day between GPS-derived data and self-reported diaries (n = 136 person-days, validation sample).

Examining individual trips, 59 (86%) of the 69 trips identified from the GPS data were found in the diary, whereas 47 (77%) of 61 trips reported in the diary were found in the GPS data.

Disagreement between GPS-based and self-reported walking trips.

Ten GPS-derived trips were not reported in a diary (four from the NHTS diary and six from the location diary). One trip was reported as a bus trip in the location diary. Although bus travel speed was expected to be higher than the selected speed range of the algorithm, heavy traffic in congested areas may have decreased actual speeds. Two trips were associated with transfers between travel modes. According to the protocol for the NHTS diary, participants were supposed to fill in a chain of subsequent travel modes in one row when they changed the travel mode of a trip toward the same destination, such as "walk-bus" or "car-bus-walk" (Fig. 2A). The algorithm identified walking activities within these trip chains, but they were classified as unmatched trips because we could not identify the exact start and end times of walking trips from the diary. The remaining seven trips were unreported. Five of them happened just before or after vehicle trips (Fig. 2B). When the participant's main mode of travel was by car, participants tended to not report walking trips to and from parking lots (Table 4).

Examples of disagreements between GPS-based and self-reported walking trips.
Disagreement between GPS-based walking trips and self-reported trips in NHTS and location diaries (n = 136 person-days, validation sample).

Fourteen diary-reported trips (nine from the NHTS diary and five from the location diary) that lasted more than 5 min were not identified with the GPS algorithm. Using the time information in the GPS data, we examined the GPS activity during the time of day each trip was reported. Five of these diary-reported trips were located in participants' residences, and no evidence of movement was detected. Presumably, respondents did not take the device with them during the reported trip, or the diary-reported time might be incorrect. The speeds of seven trips were higher (Fig. 2C) or slower (Fig. 2D) than the selected speed range. Graphically, the distance gap between GPS points in Figure 2C is larger than the gap observed in a regular walking trip (Fig. 2B). The speed for the three high-speed cases ranged from 8.6 to 10.6 km·h−1, respectively. Meanwhile, the distance gap in Figure 2D is smaller than the gaps in a regular walking trip. The speed for the four low-speed cases was between 1 and 2 km·h−1, except 1 case that was less than 1 km·h−1. Two other cases showed inconsistent speeds during the reported trip, so only parts of these trips met the selected speed range, but other parts did not.


Previous studies, with predefined pedestrian routes or activity schedules, have demonstrated that the combination of GPS and accelerometer data can predict more than 90% of activity modes accurately (27). Le Faucheur et al. (16) also reported that 90% of bouts can be classified correctly as either walking or resting with only GPS data. This study developed and tested algorithms to identify walking trips from GPS data by examining their agreement with two self-reported travel diaries. Under free-living conditions, the algorithm we selected was able to identify a high proportion of walking trips. Using an independent validation sample, a substantial level of agreement in the number and duration of walking trips between data recorded by the GPS and self-reported in both travel diaries. Examining individual trips, 86% of GPS-derived trips and 77% of diary-reported trips in both travel diaries were correctly identified as walking trips under free-living conditions.

The algorithm we developed to identify walking trips from GPS data can be improved in several ways. First, 9 of 14 cases of the "GPS-derived trips not reported on the diary" were caused by the cutoff of the speed criterion. In free-living conditions, people can engage in very slow or brisk walking and frequently make short stops (5), but it is difficult to identify those behavioral patterns with only GPS information. If the speed range was relaxed, the number of cases of the second type of disagreement, diary self-reported trips not identified with GPS, will decrease. However, the first type of disagreement, GPS-derived trips not reported in the diary, will increase as the algorithm now identifies more nonwalking trips than reported. On the basis of our results, and the results of previous studies, it may be desirable to use accelerometers with a pedometer function concurrently, which would obviate the cutoff point of the speed criterion and increase the accuracy in identifying the type of activity (1,27).

We set the Foretrex 201 GPS units to record positional coordinates at 60-s intervals. However, setting the interval to 60 s had an effect on the collection of complete information on walking trips in free-living conditions because participants frequently made short trips. Some very short walking trips, like when people move from one building to another, will remain undetected because the trips are completed before the time and position are detected by the GPS unit. To better detect characteristics of outdoor walking trips, the GPS unit could be set to record at shorter intervals, such as 10 or 20 s. Regardless of how frequently GPS units record points, some very short walking trips will remain undetected because they may be completed before records of time and position begin to be stored especially when participants move from indoors to outdoors. As technology improves, this issue should diminish.

Another possible source of inaccuracy was the respondents' lack of compliance with the protocol. Participants might have waited until the end of the day to fill out the diary. In addition, participants missed reports of walking trips to or from parking areas. There have been concerns about the unreliable nature of the data reported in diaries. Because respondents' perceptions of travel may differ from what a surveyor is attempting to collect, short trips and walking trips are often underreported in travel surveys (18). The accuracy of self-reported travel activity is questionable because respondents have a tendency to skip certain types of trips over the diary period in multiday travel diaries (14). As the number of days the diary is kept increases, the percentage of respondents reporting no travel for an entire day increases, while the number walking trips reported decreases. In addition, some have questioned the usefulness of self-reported diaries in providing specific and accurate information to support models that simulate individual-level behavior in space (4,12).

Another source of error from the lack of compliance with the protocol is that participants might have forgotten to wear the GPS units. The GPS units recorded the locations of an average of 231 min of signals per day in the calibration phase compared with an average of 166 min·d−1 in the validation phase. In the calibration phase, 31 of 41 diary-reported trips had more than 5 min of consecutive GPS data. In the validation phase, 35% of the diary-reported trips were selected to test the accuracy of the algorithms in the validation phase. The shorter duration of GPS records and the lack of consecutive GPS data in validation phase suggest that many participants in the validation phase might not have fully complied with the study protocol, so they probably missed or did not properly wear the GPS unit during some study days.

Finally, the protocols for the NHTS diary, including filling in a chain of subsequent travel modes in one row, made it difficult to identify accurate start and end points of walking trips. Reporting travel mode in separate rows may improve the ability of detecting walking trips accurately. Furthermore, the format of the NHTS and location diaries did not distinguish walking from jogging; therefore, we applied a relatively wide range of speed criterion. To obtain more specific activity types of information, it would be useful to distinguish walking from jogging in the diary.

A limitation of this study is the relatively small sample size that makes it difficult to conduct subgroup analyses. Because walking speeds are likely to differ by age and gender, identifying population-specific criteria will be an important topic when larger samples are available. Another limitation is that there was no "gold standard" against which to compare GPS-derived walking trips. The diary in the calibration phase was filled out by study staff, so we assumed it was accurate. But this is far from being a gold standard. As an alternative, during the calibration phase of a study, trained participants may be instructed to follow specific walking protocols regarding speed, direction, duration, and route. This will ensure higher-quality calibration data.


This study developed an algorithm to identify walking trips from GPS data. Consistent with prevailing vehicle-based GPS applications in travel behavior research, identifying walking trips from GPS data involved a two-step process. First, various algorithms were tested to identify the best algorithm using a calibration sample. Once identified, the best algorithm was used to identify walking trips using a separate, independent sample. Our results showed that the algorithm was capable of identifying walking trips that lasted more than 3-5 min in free-living conditions from the GPS. The algorithm will be useful for those who are interested in walking activities and their contextual information by answering questions such as where, when, how long, and how frequently walking trips occur.

Further research can determine whether the ability to detect walking accurately can be improved by having participants wear an accelerometer concurrently, reducing the recording interval on the GPS units, and checking respondents' compliance with the protocol. Improvements in GPS technology are likely to increase the ability to detect shorter-duration trips.

Preparation of this article was funded by the Robert Wood Johnson Foundation through its national program Active Living Research.

The results of the present study do not constitute endorsement by the American College of Sports Medicine.


1. Barnett A, Cerin E. Individual calibration for estimating free-living walking speed using the MTI monitor. Med Sci Sports Exerc. 2006:38(4):761-7.
2. Bassett DR, Mahar MT, Rowe DA, Morrow JR. Walking and measurement. Med Sci Sports Exerc. 2008;40(7 suppl):S529-36.
3. Bohannon R. Comfortable and maximum walking speed of adults aged 20-79 years: reference values and determinants. Age Ageing. 1997;26:15-9.
4. Chung E, Shalaby A. A trip reconstruction tool for GPS-based personal travel surveys. Transp Plann Technol. 2005;28(5):381-401.
5. Clifton K, Krizek K. The utility of the NHTS in understanding bicycle and pedestrian travel. In: Proceedings of the National Household Travel Survey Conference; 2004: Washington (DC). p. 1-18.
6. Conover WJ. Practical Nonparametric Statistics. 2nd ed. New York (NY): John Wiley & Sons; 1980. 592 p.
7. Corder K, Brage S, Ekelund U. Accelerometers and pedometers: methodology and clinical application. Curr Opin Clin Nutr Metab Care. 2007;10:597-603.
8. Du J, Aultman-Hall L. Increasing the accuracy of trip rate information from passive multi-day GPS travel datasets: automatic trip end identification issues. Transp Res Part A Gen. 2007;41(3):220-32.
9. Duncan M, Mummery K. GIS or GPS? A comparison of two methods for assessing route taken during active transport. Am J Prev Med. 2007;33(1):51-3.
10. Edward G, Robert G, Carl C, David W, Narayan V. Relationship of walking to mortality among US adults with diabetes. Arch Intern Med. 2003;163:1440-7.
11. Elgethun K, Fenske RA, Yost MG, Palcisko GJ. Time-location analysis for exposure assessment studies of children using a novel global positioning system instrument. Environ Health Perspect. 2003;111(1):115-22.
12. Elgethun K, Yost MG, Fitzpatrick CT, Nyerges TL, Fenske RA. Comparison of global positioning system (GPS) tracking and parent-report diaries to characterize children's time-location patterns. J Expo Sci Environ Epidemiol. 2007;17(2):196-206.
13. Franzblau A. A Primer of Statistics for Non-Statisticians. Oxford (UK): Haircourt, Brace; 1958. 150 p.
14. Golob T, Meurs H. Biases in response over time in a seven-day travel diary. Transportation. 1986;13(2):163-81.
15. Landis JR, Koch G. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159-74.
16. Le Faucheur A, Abraham P, Jaquinandi V, Bouyé P, Saumet JL, Noury-Desvaux B. Study of human outdoor walking with a low-cost GPS and simple spreadsheet analysis. Med Sci Sports Exerc. 2007;39(9):1570-80.
17. Li H, Guensler R, Ogle J. Analysis of morning commute route choice patterns using global positioning system-based vehicle activity data. Transp Res Rec. 2005;1926:162-70.
18. Murakami E, Wagner DP. Can using global positioning system (GPS) improve trip. Transp Res Part C Emerg Technol. 1999;7:149-65.
19. Murphy M, Nevill A, Neville C, Biddle S, Hardman A. Accumulating brisk walking for fitness, cardiovascular risk, and psychological health. Med Sci Sports Exerc. 2002;34(9):1468-74.
20. Polzin, SE, Chu X. Public Transit in America: Results from the 2001 National Household Travel Survey. Department of Transportation (FL): National Center for Transit Research Center for Urban Transportation Research; 2005. 43 p.
21. Rainham D, Krewski D, Mcdowell I, Sawada M, Liekens B. Development of a wearable global positioning system for place and health research. Int J Health Geogr. 2008;7:59.
22. Rodriguez DA, Brown AL, Troped PJ. Portable global positioning units to complement accelerometry-based physical activity monitors. Med Sci Sports Exerc. 2005;37(11 suppl):S572-81.
23. Schantz P, Stigell E. A criterion method for measuring route distance in physically active commuting. Med Sci Sports Exerc. 2009;41(2):472-78.
24. Schneider R, Patten R, Toole J. Case study analysis of pedestrian and bicycle data collection in U.S. communities. Transp Res Rec. 2005;1939:77-90.
25. Stopher PR, FitzGerald C, Xu M. Assessing the accuracy of the Sydney household travel survey with GPS. Transportation. 2007;34:723-41.
26. Stopher P, FitzGerald C, Zhang J. Search for a global positioning system device to measure person travel. Transp Res Part C Emerg Technol. 2008;16(3):350-69.
27. Troped PJ, Oliveira MS, Matthews CE, Cromley EK, Melly SJ, Craig BA. Prediction of activity mode with global positioning system and accelerometer data. Med Sci Sports Exerc. 2008;40(5):972-8.
28. Wolf J, Schonfelder S, Samaga U, Oliveira M, Axhausen KW. Eighty weeks of global position system traces. Transp Res Rec. 2004;1870:46-54.
29. Wolf J. Defining GPS and its capabilities. In: Hensher, D, editor. Handbook of Transport Geography and Spatial Systems. Oxford (UK): Elsevier; 2004. p. 411-31.


Supplemental Digital Content

©2011The American College of Sports Medicine