KEY WORDS: Physical activity, activity monitors, energy expenditure, children, validity, reliability.

Physical activity appears to have a pervasive effect on health among adults and children (Baranowski et al., 1992). Many of the risk factors for coronary artery disease, hypertension, non-insulin dependent diabetes and osteoporosis seem to arise in childhood and early adolescence (Williams et al., 1981). Several studies have suggested that physical activity in childhood is a determinant of physical activity in adulthood (Dennison et al., 1988; Kuh and Cooper, 1992). Scientific evidence supports the hypothesis that moderate and high levels of physical activity are recognized as behaviours that lead to positive health benefits (Blair et al., 1989). In a longitudinal study performed over a period of 20 years physical activity behaviour was reported for 230 healthy men and women between 13 and 33 years of age (Kemper et al., 2002). Women increased participation in sport and light physical activities other than sport (4-7 METs) and decreased participation in moderate activities (7-10 METs) with no change in the heavy ones (>10 METs). Male subjects increased participation in light and moderate sport activities but decreased participation in heavy activities over time. Although participation in physical activities other than sport decreased, participation in sport activities remained the same.

In order to better understand the relation between physical activity and health more accurate and less restrictive methods for measuring habitual physical activity and energy expenditure of men, women and children are needed (Montoye et al., 1996). Valid and reliable measures of physical activity are needed in the conduct of studies designed to develop effective programmes for promotion of physical activity (Weston et al., 1997). Even though significant steps have been made the development of valid and reliable instruments for assessing physical activity in children and youth remains an important field of interest and scientific concern (Pratt et al., 1999).

The assessment of the behavioural and the physiological aspects of physical activity resulted in the development of different measures (Baranowski et al., 1993). Self-reports have a number of advantages over other measures, but they also have limitations. Those limitations are evident when self report measures are used with children. The major sources of error identified are: the human cognitive processes, the definition of the desired variables, inadequate length of assessment and failure to account for weekday versus weekend and seasonal variations (Baronowski, 1988; Cale, 1994). Language and cultural differences may also impose additional constraints when assessing physical activity in population groups that differ from the ones that the instruments were originally validated for. Self-report methods are considered a cost effective assessment approach that is convenient to administer and feasible when testing large populations (Sallis et al., 1993).

To reduce the errors associated with self report methods in children, objective methods such as motion sensors have been developed. The Computer Science and Applications Inc. (MTI/CSA) uni-axial activity monitor (WAM 7164; Computer Science and Applications Inc., Shalimar, FL) is one example (Ekelund et al., 2001). The MTI/CSA monitor provides a precise tool for measuring changes in acceleration in laboratory settings (Metcalf et al., 2002) and can be useful in a field situation (Sirard et al., 2000). The validation of the MTI/CSA certifies this monitor as valid, reliable and useful device for the assessment of physical activity in children (Trost et al., 1998; Puyau et al., 2002). Its small size and lightness creates an advantage when compared to other direct methods of estimating physical activity such as heart rate monitoring. Other advantages that accelerometers have when compared to other methods of measuring physical activity include storage of movement data for long periods of time, objective estimation of frequency, intensity and duration of physical activity and continuous recording. The continuous recording of physical activity data for future recall excludes problems related to subjective recall that physical activity questionnaires demand (Basset et al., 2000).

Despite the advantages of accelerometers, the energy cost of an activity that includes muscular activity resulting from isometric contractions, from movements of the upper body, from additional load bearing (carrying, lifting, and pushing), and from movement on inclined or soft surfaces will not result in an increase in the number of counts recorded by any uni-axial accelerometer device (Bouten et al., 1994; Hendelman et al., 2000). Additionally accelerometers cannot be used while swimming or during any other activities where water is involved, and their effectiveness is limited in recording activities such as cycling and weight lifting (Basset et al., 2000). Most motion sensors are sensible to velocity changes but not to gradient changes. This lack of sensitivity may be important in laboratory settings but in field studies variations in duration, frequency and intensity which are the most important components of physical activity are still recorded (Montoye et al., 1983).

Four days of measurement, including weekdays and weekend days, should provide acceptable correlations ensuring a representative measure of a child's habitual physical activity in a large field based study (Janz, 1994; Janz et al., 1995; Trost et al., 2000). In the present study children wore the MTI/CSA monitors for 7 consecutive days during the two periods of measurement in order to achieve greater reliability in recording physical activity (Trost et al., 2000; Matthews et al., 2002).

The aim of the present study is to report on the reliability of three questionnaires 3DPAR (Bouchard et al., 1983), 4BY1RPAQ (Cale, 1993) and PARQ (Avgerinos, 2000) when used to record physical activity of Greek High School children. The validity of these instruments was assessed by using the MTI/CSA accelerometer as the criterion standard.

Subjects
Forty children (23 boys and 17 girls) ranging in age from 13 to 14 years (mean 13.73 ± 0.80 years) participated in this study. Subjects were randomly selected and recruited from 7 high schools with the help of the physical education teachers who worked at the schools. The children and their parents were informed about the scope and the procedure that would be followed in this study and all parents signed an informed consent form.

Both body mass and height were recorded for all subjects. Height was measured without shoes by using a wall - mounted tape measure, and body mass was measured using a digital laboratory scale (Seca, Model 770, Olney, MD). Body mass index (BMI) was computed as weight in kilograms divided by height in square meters.

Instruments
Motion Detector MTI/CSA 7164
The MTI/CSA (7164) activity monitor is a small (5.1 x 3.8 x 1.5cm) lightweight (43-45g) uni-axial activity monitor. It is designed to detect acceleration ranging in magnitude from 0.05 to 2.0G with a frequency response from 0.25 to 2.50 Hz (CSA Inc., Shalimar, FL 1995). These parameters allow the detection of normal human motion while rejecting high frequency motion encountered outside these ranges. The filtered acceleration signal is digitized and the magnitude is summed over a user-specified epoch interval. At the end of each interval, the summed value or activity "count" is stored in memory. The stored data can be downloaded to a computer and the integrator can be reset. For the current study a one minute time interval was used. Further technical specification and performance properties have been described elsewhere (Janz, 1994; Melanson and Freedson, 1995; Freedson et al., 1998; 2000; Nichols et al., 2000; Swartz et al., 2000; Ekelund et al., 2000; 2001).

The MTI/CSA accelerometer has been well validated in both children and adolescents against a wide range of outcomes (Freedson et al., 1998; Trost et al., 1998; Freedson and Miller, 2000; Welk et al., 2000; Ekelund et al., 2000; 2001; Brage, et al., 2003). It has been validated against energy expenditure measured by indirect calorimetry and it was found to be a valid tool for quantifying energy expenditure in children and adults during treadmill running and walking. Correlations ranging from r = 0.50 to r = 0.74 have been reported between the MTI/CSA monitor and heart rate telemetry of children in field settings (Janz, 1994). Additionally the MTI/CSA monitor provided more accurate estimations of energy expenditure when compared to TriTrac and other accelerometers (Welk, et al., 2000). In studies with children, a significant correlation was observed between MTI/CSA activity counts and all energy expenditure estimates using the Doubly Labeled Water method (Ekelund et al., 2001).

Three-day physical activity record
This three-day activity record divides each of the three days (one must be a weekend day) into 96 periods each 15 minutes long. The responder records his/her energy expenditure for each 15 minute period using a scale from 1 (sleep) to 9 (vigorous physical activity and sport). The values of this scale correspond to a range of 1.0 to 7.8 METs and higher. These categories are explained to the participant in material given to him or her for personal use during the completion of the 3DPAR. Approximate median energy cost for each of the nine categories in (Kcal·kg^-1·15 min^-1) was used to compute the daily energy expenditure for each individual in kcals (Bouchard et al., 1997). The reliability of this questionnaire has been tested for both children and adults (ICC = 0.86 - 0.95, p < 0.01) and its validity has been also reported for children (r = 0.80, p<0.01) and adults (r = 0.54, p < 0.01) (Bouchard et al., 1983).

The four by one-day recall physical activity questionnaire
This interviewer administered self-report measure was designed for children aged 11 years and older and is used to gather four days of activity information (2 school days and two weekend days). It consists of two forms: school day and weekend day, the form is segmented into parts: morning, afternoon and evening and contains checklists of activities. The questionnaire measures four dimensions of physical activity: a) physical activity at school, b) sport at school, c) physical activity during leisure time and d) sport during leisure time. Additionally the questionnaire provides an objective scoring system and measure of physical activity in terms of: a) average daily energy expenditure (Kcal· kg^-1·day^-1), b) time spent on moderate physical activity, c) time spent on hard and very hard activity and d) bouts of "huff" and "puff" activity. The final score assigns the responder to one of the following categories: very inactive, inactive, moderately active and active. The reliability and validity of this questionnaire have been tested for 11 to 14 year-old children and have been reported as r = 0.62 (p < 0.05) (reliability) and r = 0.61 (p < 0.01) (validity) (Cale, 1994).

Physical activity and lifestyle questionnaire
This self-administered questionnaire is designed to assess the physical activity of youngsters, ages 11 to 18 years. The questionnaire is made up of two parts: Part 1 records participation in physical activity during leisure time in which the responder must record how frequently s/he participates in any of the 27 physical activities listed in the questionnaire. Part 2 assesses the responder's participation in physical activities during the last 7 days. PALQ can be answered in 15 to 25 minutes depending on the age of the responder. The scoring procedure results in a total score of energy expenditure for each responder that can be calculated using the Compendium of Physical Activities (Ainsworth et al., 1993). Subjects are assigned to one of four categories according to their score: very inactive, inactive, moderately active and active. Additionally daily time of participation in vigorous and moderate physical activities and sport can be calculated from the data collected. Only data collected for the second part was used in this study.

Selection of questionnaires used in this study
No self-report and/or interviewer-administered instruments that assess physical activity have been validated for use with Greek High School children (Avgerinos, 2000). The researchers adapted the 3DPAR and the 4BY1RPAQ questionnaires in order to use them with Greek High School children since both questionnaires have been reported as validated and reliable instruments in studies assessing physical activity in children.

The three questionnaires used in this study assess physical activity by estimating energy expenditure either in METs or in Kcal. Additionally, the content and the format of the questions included in 3DPAR and 4BY1RPAQ was modified to be relevant to the cultural characteristics of Greek children. The adaptation from English to Greek language did not require any changes in either the number or meaning of the questions (both questionnaires). Additionally, by using 3DPAR and 4BY1RPAQ we would able to use and provide data, for Greek children, that is comparable with data from studies conducted in other countries and assess how well these instruments can be adapted and used in Greece.

4BY1RPAQ uses the Compendium of Physical Activities (Ainsworth et al., 1993), which is also used by PALQ, a new instrument that has been designed to assess the physical activity of 12-18 year old Greek students. PALQ is the first questionnaire devised in Greece that provides native researchers and Physical Education instructors with the means to assess the habitual physical activity of Greek high school students.

Standard methods were used to translate and adapt the 4BY1RPAQ and the 3DPAR questionnaires. Two pilot studies were conducted in order to: a) check the childern's understanding of the questions and the material included in the translated versions of the two questionnaires, and b) determine the time required to read and complete them. The subjects in these pilot studies were Greek high school students, (n_-1 = 6 boys and 4 girls aged 13.57 ± 0.70 years and n₂ = 8 boys and 7 girls aged 13.64 ± 0.70 years). Where the subjects had difficulties in understanding the material given to them, additional information was included in the final versions of the questionnaires.

Study Design
The study was designed to compare the criterion method, the MTI/CSA, against a single-interviewer administered questionnaire and two self-report methods of assessing physical activity. The study entailed 14 days of children's involvement. In order to provide a measure of reliability and/or behaviour stability, each questionnaire was administered twice and for a two week interval. Additionally, the subjects wore the MTI/CSA accelerometer twice for seven consecutive days (2 non consecutive weeks) during the same days that they responded to the questionnaires. Activity monitoring data was collected for 14 days.

All interviews were conducted in an environment familiar to the subject (school or home). Subjects were encouraged to maintain their daily physical activity schedule and in case of illness the study was discontinued. The MTI/CSA accelerometer was placed in a carrying pouch and participants were shown how to place the pouch on a belt at waist level on the right anterior axillary line. The investigator checked the functionality of the monitor every morning during the period of the study. Consistent and proper placement was emphasized, and subjects were told to wear the monitor at all times (day and night), even while sleeping, and to remove it only if the monitor would get completely wet, such as when showering or swimming. The investigators trained the subjects how to attach and detach the accelerometer. During both pilot studies, subjects would repeatedly forget to put on the MTI/CSA monitor as they were getting ready to go to school in the morning. This observation was the reason behind the decision to advise the children to continuously wear the MTI/CSA monitor. As a result, MTI/CSA data was collected for a period of 24 hours each day, which was also the case for the data from the questionnaires used in this study. Compliance was monitored by investigators every morning as the children arrived at school. The MTI/CSA was initialized according to the manufacturer's specifications. The epoch interval was set for 1 minute. At the end of the 7-day recording period, the investigator used a reader interface unit to download the MTI/CSA data (counts) to a desktop computer.
The data was then stored in an Excel file for further analysis.

The 3DPAR was completed by the children and returned to the investigator at the end of the third day. The 4BY1RPAQ was administered by the investigator for two weekdays and two weekend days and referred to the physical activity undertaken by the subject the previous day. The third questionnaire (PALQ) was completed just after the end of the seven-day period during which the subjects wore the MTI/CSA accelerometer. Data collection was completed within a period of one month since changes in weather conditions over long periods of time could affect the children's habitual physical activity. The possible effects due to weather changes were observed during the pilot study. The above procedure ensured that the questionnaires were recorded during identical periods of time, their guidelines were satisfied and activity recall from the completion of another questionnaire was minimized. Moreover, this design enabled the researchers to comparatively evaluate the data of all the measures of physical activity used in this study. (However, comparative evaluation is not discussed in this paper).

Statistical Analysis
The physical characteristics of the children were summarized using descriptive statistics. For reliability, intraclass correlation coefficients (ICC) were calculated for the repeated administrations of the MTI/CSA, the 3DPAR, the 4BY1RPAQ and the PALQ, using appropriate two-way ANOVA models (Baumgartner, 1989). A two-way ANOVA with repeated measures (week x day) was performed in order to determine differences between scores obtained during the two measurements for MTI/CSA, 3DPAR and 43BY1RPAQ. Post-hoc multiple comparisons were performed using the Bonferroni test. Significance for all parts of the analysis was determined at an alpha level of p<0.05. A paired Student's t-test was performed for the two scores obtained with the PALQ. Two additional measures of reliability were also computed for the MTI/CSA, the 3DPAR, the 4BY1RPAQ and the PALQ: the typical error (Hopkins, 2000) and the limits of agreement (LOA) (Bland and Altman, 1986). Pearson product moment correlation coefficients between the MTI/CSA and the three questionnaires were used to assess their validity.

Participants Physical Characteristics
Boys' (n = 23) body mass (mean ± SD) was 58.7 ± 12.2 kg, their height was 1.66 ± 0.11m and their BMI was 21.3 ± 3.54. Girls' (n = 17) body mass (mean ± SD) was 52.1 ± 10.8 kg, their height was 1.60 ± 0.07m and their BMI was 20.3 ± 3.3. From the BMI readings obtained in this study, boys resulted between the 75^th and 85th percentile distribution and girls resided between the 50th and the 75^th percentile. (National Center for Health Statistics, USA, 2000), both groups being between the normal range (Himes and Deitz, 1994).

The mean ± SD for the 4 instruments obtained during the first and second measurement period is presented in Table 1. Original 4BY1RPAQ and PALQ data was transformed from METs to Kcal.

Reliability and Consistency
MTI/CSA
Average MTI/CSA scores (counts·min-1) for the first (A Data) and the second measure (B Data) were presented in Figure 1. The ICC for average MTI/CSA counts·min^-1 across the two weeks of measurement was 0.52 (p <0.05). The ICCs for the separate days were: Saturday ICC = - 0.49, Sunday ICC = 0.66, Monday ICC= 0.42, Tuesday ICC = 0.55, Wednesday ICC = 0.48, Thursday ICC = 0.65 and Friday ICC = 0.35. The typical error for MTI/CSA between the two weeks of measurement was 1099.05counts·min^-1 and the respective LOAs were 38.84 ± 3139.68 counts·min^-1. The typical error and the LOAs for the separate days were for: Saturday 265.68 and 55.68 ± 758.97 counts·min^-1, Sunday 129.15 and -33.82 ± 368.96 counts·min^-1, Monday 121.64 and -11.03 ± 347.48 counts·min^-1, Tuesday 259.10 and -28.58 ± 740.17 counts·min^-1, Wednesday 207.78 and 41.45 ± 593.58 counts·min^-1, Thursday 221.78 and -26.98 ± 633.56 counts·min^-1 and Friday 299.32 and 42.03 ± 835.08 counts·min^-1. No significant (week x day) interaction was reported (F1,17 = 0.003, p > 0.05), and no significant main effect for the factors week, (F1,17 = 0.009, p > 0.05) and day (F1,17 = 0.577, p > 0.05).

3D Physical Activity Record.
The ICC for the two 3day periods of measurement was (ICC = 0.97, p < 0.01). When ICC was calculated for each day of measurement separately, weekdays gave higher ICC readings (weekdays 1 and 2 ICC = 0.97, p <0.01) than the weekend day (ICC =0.88, p < 0.01). The typical error and the LOAs for 3DPAR between the two 3day periods of measurement were 382.51 and -375.30 ± 1092.72 Kcals respectively. The typical errors for weekend day, weekday 1 and weekday 2 were respectively 276.36, 119.78 and 131.48 Kcals. The corresponding LOAs were -230.60 ± 789.49 Kcals, -66.12 ± 342.19 Kcals and -78.58 ± 375.60 Kcals. No significant (week x day) interaction was reported (F_1,20 = 2.746, p > 0.05). The total scores between the 2 weeks of measurement were not significantly different (F_1,20 = 3.178, p > 0.05). A significant main effect was reported for the factor: day (F_1,20 = 9.172, p<0.01). A Bonferroni post-hoc test revealed that the weekend day scores (Sunday) of the 2 weeks of measurement were significantly different (Figure 2).

4BY1 Recall Physical Activity Questionnaire
The ICC for the two 4day periods of measurement was (ICC = 0.70, p < 0.01). When ICC was calculated for each day of measurement separately weekdays had higher correlation coefficients (weekday 1 ICC = 0.83, p < 0.01, & weekday 2 ICC = 0.89, p < 0.01,) than weekend days (Saturday ICC = -0.30, p > 0.05, and Sunday ICC = 0.47, p > 0.05). The typical error and the LOAs for 4BY1RPAQ between the two 4day periods of measurement were 7.94 and -4.37 ± 22.67 METs respectively. The typical errors for each of the weekend days were equal to 4.77 METs and for each of the two weekdays were respectively 1.25 and 2.25 METs. The corresponding LOAs were 0.10 ± 13.63, -3.25 ± 13.64, -0.81 ± 3.56 and -0.40 ± 6.44 METs. No significant (week x day) interaction was reported (F_1,20 = 0.020, p > 0.05). No differences were reported between the total scores of the 2 weeks of measurement (F_1,20 = 10.108, p > 0.01). A significant main effect was reported for the factor: day (F_1,20 = 4.624, p < 0.05). A Bonferroni post-hoc test revealed that Sunday scores of the 2 weeks of measurement were significantly different.

Physical Activity and Lifestyle Questionnaire
The ICC for the two periods of measurement was (ICC = 0.52, p < 0.05). The typical error and the limits of agreement, for PALQ, between the two periods of measurement were respectively 2.39 and -1.88 ± 6.82 METs.The physical activity score for the first week was significantly different from the score of the second week (t = 2.547, p < 0.05).Average energy expenditures (METs) recorded with 4BY1RPAQ per day and PALQ for the first (A Data) and the second measure (B Data) were presented in Figure 3.

Validity of the Physical Activity Instruments
MTI/CSA vs. 3D Physical Activity Record.
The correlation between 3DPAR (kcal) and MTI/CSA (counts.min^-1) for the respective days was moderate (r = 0.63, p<0.01) (Table 2). The correlations of each of the three individual days between the 3DPAR and the MTI/CSA monitor were significant for weekend day 2 (r = 0.56, p<0.01) and for the weekday 2 (r = 0.64, p<0.01).

MTI/CSA vs. 4BY1RPAQ Recall Physical Activity Questionnaire
The average correlation between the 4 days of 4BY1RPAQ (METs) and the MTI/CSA counts·min^-1 was moderate (r = 0.62, p < 0.01, Table 2). The correlations of each of the four days between the 3DPAR and the MTI/CSA monitor were also significant for both weekend days (Saturday r = 0.38, p < 0.01 and Sunday r = 0.54, p < 0.01) and weekdays (Monday r = 0.46, p < 0.01 and Tuesday r = 0.65, p < 0.01).

MTI/CSA vs. Physical Activity and Lifestyle Questionnaire
The average correlation between the 7 days of PALQ (METs) and the respective days of MTI/CSA (counts·min^-1) was moderate (r = 0.53, p < 0.01, Table 2).

• The PALQ demonstrated a moderate reliability (0.52) and validity (0.53) in recording physical activity.
• A relatively high correlation was observed between the MTI/CSA and 3DPAR and a moderate correlation was observed between MTI/CSA and the 4BY1RPAQ tested in this study.
• Only a combination of the available instruments would be able to respond to the interpersonal and intrapersonal variability when assessing physical activity in children and adolescents. Self-report instruments and accelerometers are probably able to quantify only gross fluctuations in physical activity.
• All 4 instruments used in this study were valid and reliable in recording physical activity when used with children, since the instruments were able to detect changes in physical activity and the respective energy cost.