Share this article on:

Front Squat Data Reproducibility Collected With a Triple-Axis Accelerometer

Caruso, John F; Olson, Nathan M; Taylor, Skyler T; McLagan, Jessica R; Shepherd, Catherine M; Borgsmiller, Jake A; Mason, Melissa L; Riner, Rebekah R; Gilliland, Laura; Grisewold, Shawn

The Journal of Strength & Conditioning Research: January 2012 - Volume 26 - Issue 1 - p 40-46
doi: 10.1519/JSC.0b013e31821d5ed7
Original Research

Caruso, JF, Olson, NM, Taylor, ST, McLagan, JR, Shepherd, CM, Borgsmiller, JA, Mason, ML, Riner, RR, Gilliland, L, and Griswold, S. Front squat data reproducibility collected with a triple-axis accelerometer. J Strength Cond Res 26(1): 40–46, 2012—The purpose of our study was to assess data reproducibility from 2 consecutive front squat workouts, spaced 1 week apart, performed by American college football players (n = 18) as they prepared for their competitive season. For each workout, our methods entailed the performance of 3–6 front squat repetitions per set at 55, 65, and 75% of subject's 1 repetition maximum (1RM) load. In addition, a fourth set was done at a heavier load, with a resistance equal to 80 and 83% of their 1RM values, for the first and second workouts, respectively. A triple-axis accelerometer was affixed to a barbell to quantify exercise performance. Per load, the accelerometer measures peak values for the following indices: force, velocity, and power. To assess data reproducibility, inter–workout comparisons were made for 12 performance indices with 4 statistical test-retest measures: intraclass correlation coefficients, coefficients of variation (CVs), and the SEM expressed in both absolute and relative terms. Current results show that the majority of performance indices exceeded intraclass correlation (0.75–0.80) and CV (10–15%) values previously deemed as acceptable levels of data reproducibility. The 2 indices with the greatest variability were power and velocity values obtained at 55% of the 1RM load; thus, it was concluded that higher movement rates at the lightest load were the most difficult aspect of front squat performance to repeat successfully over time. Our practical applications imply lighter loads, with inherently higher rates of barbell movement, yield lower data reproducibility values.

1Exercise Physiology Laboratory, Exercise and Sports Science Program, The University of Tulsa, Tulsa, Oklahoma; and 2Athletics Department, The University of Tulsa, Tulsa, Oklahoma

Address correspondence to Dr. John Caruso,

Back to Top | Article Outline


The accurate measurement of exercise performance is continually sought by persons interested in monitoring the progress of athletes or to assess the merits of workout programs. Thus the ability of devices to measure fitness and exercise performance routinely have their data evaluated for their reproducibility, which is an index of the internal consistency or repeatability of values collected over time. Aerobic measures of data reproducibility (V̇O2max, respiratory exchange ratio [RER], etc.), obtained over multiple trials under steady-state conditions, yielded values with modest amounts of variability (10,14,15). In contrast, measures of resistive exercise performance, particularly for workouts that entail dynamic high-speed movements, typically elicit less data reproducibility (30). Current technology to measure high-speed resistive exercise performance includes accelerometers, which are specifically designed to quantify the rate of velocity change during linear barbell movements (6,18,23,27). Perhaps, the most advanced devices of this type are triple-axis accelerometers. They may be affixed to a barbell and potentially have the capability to measure acceleration in all 3 planes of motion simultaneously. Accelerometers are currently the most advanced devices to monitor an athlete's progress with respect to their resistive exercise workout performance.

Per resistive exercise set, accelerometers may measure peak force, velocity, and power. Although the reproducibility of accelerometer data has been examined (6,18,23), such studies usually occurred in laboratories with subjects who may have had little experience as competitive athletes. Data reproducibility results from such studies may thus not be applicable to the workouts done by competitive athletes, because they typically train in a more “explosive” (higher barbell velocities and heavier loads) fashion as compared with many of the subjects used in prior (6,18,23) studies. In addition, although the reproducibility of resistive exercise data was assessed previously (6,18,23), many of the prior studies did not employ standard weight training (barbells, dumbbells, etc.) equipment (16,27), which makes their results even less applicable to athletes who customarily use more conventional strength training devices. To closely monitor an athlete's training progress, or to assess the merits of workout programs, such data would be obtained from them as they trained for competition under their normal workout conditions. In addition, such data should be derived from an exercise integral to an athlete's workouts, such as a multijoint movement that engages some of the larger muscles of the body. Such exercises include the front squat.

A variation of the deep knee bend, the front squat is a supplementary movement for the Olympic lifts because both are performed with the barbell in front of the body. Unlike the classic version of the deep knee bend in which the barbell rests across the back of the shoulders, front squats are usually done at higher speeds and may thus best enhance power production for the Olympic lifts. Thus, the front squat is a good exercise to aid power development in athletes. Because of its value in training programs, assessments of data reproducibility from performance variables derived from the front squat with an accelerometer should be quantified. Although some studies compared the front and back squat exercises (6,13), few trials have focused solely on the frontal version of the exercise. To measure the repeatability of data collected over a 1-week period, our study assessed the reproducibility of performance variables obtained from the front squat exercise with a triple-axis accelerometer. Data were provided by college athletes under their normal workout conditions as they prepared for their competitive season. We believe our results will be more applicable to competitive athletes than values obtained from studies comprised of nonathletic samples (6,18,23). Data reproducibility will be assessed with 4 statistical test-retest measures. We hypothesize that our results will yield acceptable levels of data reproducibility.

Back to Top | Article Outline


Experimental Approach to the Problem

The purpose of this study is to assess the reproducibility of data provided by a triple-axis accelerometer. To make our results more applicable to coaches and JSCR readers, we obtained data from competitive varsity athletes as they performed an exercise integral to their workouts.

Back to Top | Article Outline


Study procedures first received approval for the use of human subjects by a university-based Institutional Review Board. Before their involvement, the subjects provided informed written consent to participate. The subjects were in good health and free of musculoskeletal injuries. To address the purpose of this study, our experimental approach included subjects (n = 18; mean ± SD, body mass index 31.2 ± 5.5 kg·m−2) with an average of 5 years' experience as American football players, who also were familiar with the front squat for at least 2 years before their project involvement. At the time of data collection, the subjects were varsity athletes. Data were collected over 2 consecutive front squat workouts, separated by 1 week, and entailed nearly identical protocols. Workouts occurred in a varsity weight room used exclusively to train student athletes. The exercise protocols were designed and scheduled by the university's head strength coach. The subjects performed front squats as part of their customary workout program. With a triple-axis accelerometer attached to the barbell, front squat exercise performance was quantified after each set. Four test-retest measures then quantified within-subject interworkout data reproducibility.

Back to Top | Article Outline


Both front squat workouts were part of a training mesocycle intended to increase power production. The current subjects were on a regimented exercise and dietary program prescribed by their coaches. The subjects arrived at the varsity weight room in a well-rested and hydrated state. Their dietary program included meals prepared and eaten at a varsity training table. Because of the regimented nature of their dietary program, this likely added little variability to study results. Throughout the workouts, the subjects had the opportunity to consume an electrolyte-based fluid replacement beverage ad libitum. For the current workouts, the weight room had an ambient room temperature range of 18–20°C and an average relative humidity of 35%. Thus, we do not believe that the fluid balance of our subjects led to dehydration or variations in their workout performance.

Each workout began with 25 minutes of calisthenics, conditioning drills, and power cleans so that the subjects were fully warmed up when they performed their front squat sets. Each subject performed 4 front squat sets separated by 3- to 4-minute rest periods. Spotters were present, and vocal encouragement was provided for each set. Subjects' head strength coach, who was also present at each workout, had recently overseen the performance of their front squat 1 repetition maximum (1RM). For the first workout, over consecutive sets, the subjects performed 3–6 repetitions at 55, 65, 75, and 80% of their 1RM load. An identical protocol was used for the first 3 sets of the second workout. Yet for the final set of their second workout, the subjects used a load equal to 83% of their 1RM. Thus, the difference in the final set load for the 2 workouts is a potential source of data variability.

Back to Top | Article Outline

Triple-Axis Accelerometer

A triple-axis accelerometer (Myotest Inc., Royal Oak, MI, USA) was affixed to the barbell to measure front squat performance. The mechanical characteristics of the accelerometer include a crossaxis sensitivity of 2% (25). The accelerometer was used in accordance with the manufacturer's guidelines. When placed in the trainer mode for collection at 500 Hz, the accelerometer began to obtain data at the sound of an automated beep that emanated from the device. As the beep occurred, the subjects were in the front squat starting position (Figure 1) with an unracked barbell resting atop their clavicles. After the beep, the subjects performed the prescribed number of repetitions. When done at each subject's customary pace, repetitions emphasized moving as rapidly as possible throughout the ascent. Per repetition, the subjects descended until their dorsal surface of their thighs were parallel to the floor before they began their ascent. The subjects descended to the point at which the bottoms of their thighs were parallel to the floor. Figure 2 depicts their ascent.

Figure 1

Figure 1

Figure 2

Figure 2

When the prescribed number of repetitions was completed, the subjects stood motionless until a second beep was produced by the accelerometer. The aubjects then reracked the barbell as the accelerometer concurrently displayed peak force, velocity, and power values in newtons, centimeters per second, and watts, respectively. Data from each of the 3 axes of the accelerometer were integrated to quantify performance. Per plane of motion, each axis measures acceleration or the rate of velocity change with respect to time. Velocity changes were derived by multiplication of acceleration data by the change in time. Based on a knowledge of the rate of movement at prior time points and the change in speed over given time intervals, velocity values were derived. Data from each axis were then integrated to yield a peak velocity. Force was calculated as the product of mass and acceleration; with barbell mass programmed into the accelerometer before each set. Peak force (F), velocity (V), and power (P) measurements were recorded from the 4 different loads (55, 65, 75, and 80–83% 1RM) for a total of 12 performance variables of this study.

Back to Top | Article Outline

Statistical Analyses

To assess the reproducibility of accelerometer data, we paired first workout values to the corresponding numbers from the second workout. Such within-subject pairwise comparisons occurred for each performance variable. In regard to the calculation of the current sample size, with a power level of 0.80 and an Ω2 value of 0.15, our n = 18 exceeds the minimum number required to perform our statistical analyses (17). Z-scores were first used to identify statistical outliers for all 12 performance variables. Outliers, and their paired value derived from the other workout, were omitted from further analyses. Within-subject interworkout comparisons were then made, with 4 test-retest statistical tools, to assess data reproducibility. Comparisons were made with peak values for the following performance indices: force, velocity, and power at the 3 lighter barbell loads (F55, F65, F75, V55, V65, V75, P55, P65, P75). In addition, though the 2 workouts had different barbell masses for the final sets, interworkout comparisons were also made for the heavier loads (F80–83, V80–83, P80–83). Thus, 12 performance variables were assessed for their data reproducibility with 4 statistical test-retest measures.

Data reproducibility was assessed with intraclass correlation coefficients (ICCs), coefficients of variation (CVs), and the standard error of measurement expressed in both absolute (SEM) and relative (SEM%) terms. The ICC can assess the level of agreement among paired data (4,12). The CVs were computed as the SD/mean ratio and then multiplied by 100 to express values as a percentage (14,26). Sometimes referred to as the within-subject variance, the SEM was computed as follows: SD × (1 − Pearson product moment correlation coefficient)0.5 (1,2,4). The SEM% was derived as the SEM/mean ratio and then multiplied by 100 to express values as a percentage. The ICC values range from −1 to +1; the extreme ends of that range denote high reproducibility, whereas ICCs near zero represent a large degree of data variability. Unlike the ICC, lower CV, SEM, and SEM% values denote greater data reproducibility.

Back to Top | Article Outline


No subjects were injured through their project involvement. Z-score analyses detected no statistical outliers. Raw data, with values averaged across workouts for each performance variable, appear in Table 1. In addition to the 12 performance variables, raw data for the front squat barbell loads (L55, L65, L75, L80–83) also appear in Table 1. The ranges and SDs in Table 1 reveal considerable variability. Yet it is important to realize that differences in the absolute 1RM values of our subjects was a major contributor to the Table 1 data variability and are reflected in the L55, L65, L75, and L80–83 values. The current sample included 10 lineman or linebackers, whereas 8 played the so-called “skilled” positions that demand greater ball handling and running skills. Because of the unique skill set each American football playing position demands, lineman and linebackers are usually heavier and possess greater absolute 1RM strength than do their teammates and contributed to the greater data variability seen in this study (21).

Table 1

Table 1

Data reproducibility results for the 12 variables appear in Table 2. Overall, given the study design, subjects and data collection conditions, the current results may be considered to demonstrate acceptable levels of repeatability for our front squat data. In terms of ICC results, 10 of the 12 performance variables either met or exceeded what is considered an acceptable level of reproducibility (19,20). Typically, for each statistical test-retest tool the performance indices obtained at the lightest load, such as P55 and V55, had the highest data variability. Table 2 shows that force-based indices were the most reproducible, whereas velocity and power variables generally had more variability. This suggests that within-subject interworkout barbell movement rates were perhaps the least repeatable performance features of this study. Because power was calculated as the product of force and velocity, if either of those factors had a high degree of variability by themselves, it in turn led to lower reproducibility values for our power indices.

Table 2

Table 2

Back to Top | Article Outline


Accelerometers have typically quantified general measures of physical activity (10), yet other investigations also assessed their ability to yield reproducible data over multiple resistive exercise workouts (6,18,23,27). Although results from these prior resistive exercise studies included acceptable levels of reproducibility, their data were obtained in laboratories from subjects with diverse resistive exercise training backgrounds (6,18,23,27). Thus, the applicability of such data to coaches and athletes is somewhat limited (18,27). In contrast, the opportunity to assess the reproducibility of data obtained from athletes as they perform real workouts in preparation for their competitive season is rare; yet this was precisely the purpose of this study. Despite the challenges of the design and purpose of this study, which was novel for a resistive exercise data reproducibility study, our hypothesis was affirmed. Based on criteria established from other studies (19,20,24,26), Table 2 test-retest results shows that most performance variables either met or exceeded what was considered an acceptable level of data reproducibility.

Accelerometry evoked high ICC peak power, time to peak power, and rate of power development values for bench presses done by male nonathletes (18). A bench and leg press exercise study showed that accelerometry values from multiple trials yielded acceptable levels of reproducibility; it was then implied that their values could be averaged to offer more stable variable measurements (5). The reproducibility of accelerometer data was also assessed from men who performed multiple workouts on a linear bench press machine (27). The subjects performed 3 trials at 60% of their 1RM and repeated the lifts 24 hours later (27). The length of the time between workouts was less than that for this study and, given the submaximal nature of the exercise protocol, raised the likelihood to achieve acceptable levels of data reproducibility. Power, velocity, and total displacement were measured with 3 collection devices: a single-axis accelerometer affixed to a moveable arm of the exercise machine, a video camera that recorded each trial, and a photocell equipped with a timer to measure segments of the movement (27). The results for each collection system offered high intertrial reproducibility, yet accelerometry values were slightly but significantly different when data for each workout were assessed over a full range of motion. It was concluded, that although acquisition errors and differences existed, the accelerometry data were reproducible (27).

Accelerometry data were used to predict 1RM bench press loads in men (23). Because force is the product of mass and acceleration, it was theorized that 1RM loads could be calculated with repetitions done at submaximal weights, provided maximum acceleration values were derived. With an accelerometer attached to a barbell, bench presses were performed at 5 (50, 60, 70, 80, and 90% 1RM) submaximal loads (23). Pearson product moment correlation coefficients, which compared actual 1RM values with those computed from repetitions done at each submaximal load, ranged from 0.89 to 0.97 (23). The standard error of the estimates ranged from 4.5 to 19.1 kg for the 5 submaximal weights, with an average value of 12.0 kg when data from all loads were pooled (23). It was concluded that accelerometry was a better predictor of true 1RM bench press values from repetitions done at the higher (80 and 90% 1RM) submaximal loads. Thus, in agreement with current results, which saw more variability from sets done at 55% of the 1RM load, prior accelerometry studies noted that data reproducibility is compromised at a lighter resistance (23,27).

Many current performance indices met or exceeded what was considered acceptable levels of data reproducibility (19,20,24,26); the 2 exceptions (P55, V55) relate to the ability to consistently repeat higher rates of exercise movement over time. Perhaps, perceptions of maximum effort lifts done over multiple trials, such as those for the current subjects, led them to focus more on the absolute load rather than on the rate at which repetitions occurred. In addition, as our data were collected, the subjects may have put forth less mental and physical effort for sets done at the lightest load, which may have added to the P55 and V55 variability. The ability to repeat exercise movement rates over time was assessed in a study whose main purpose was to examine the reproducibility of data obtained from men (30). With data collected from the subjects twice over a 10-day period, peak torque, movement time, and rate of torque development were measured from dynamic and isometric repetitions. The dynamic actions entailed both movement against no extra resistance and an added load equal to 10% of the subject's maximum voluntary contraction (30). With both types of dynamic repetitions done at high rates of movement, results consistently showed more data variability for the dynamic vs. the isometric, repetitions. It was implied that data collection and testing methods should be refined with respect to dynamic repetitions and speaks to the difficulty to consistently repeat high rates of exercise movement over time (30).

Table 2 shows the test-retest results for paired values from consecutive front squat workouts spaced 1 week apart. The level of agreement among paired values may be quantified with ICC (3,12). A ratio of the true variance of interest to error variance obtained from measurements (22), the ICC does not inflate the strength of agreement among paired values or show bias toward small samples (15). An early paper claimed that ICC values of 0.75–0.80 denoted an acceptable level of reproducibility (24). More recent exercise studies (19,20), whereby repetitions were performed on a fitness machine over a single plane of motion for a limited number of repetitions, adopted these ICC values as a benchmark for acceptability. With this criterion, our results show that the majority of current performance indices exceeded the ICC values deemed acceptable. The ICC values that did not (P55, V55) relate to the ability to replicate high exercise movement rates over time, because this again appears to be the most difficult aspect to reproduce, even among athletes.

Akin to the ICC, the level of distribution and dispersion for paired data may be assessed with the CV (14). Test-retest measures are impacted by the study design, subject characteristics, and data collection conditions, yet CV values of 10–15% were deemed typical for biological systems (26). Current force indices were below the threshold deemed typical (26). Recent trials (16,28,29) had CV values similar to the force indices in Table 2. Interworkout CV values of 3.9–5.9 resulted from isokinetic knee extension and plantar flexion repetitions, because higher movement rates yielded more variability (28,29). Data reproducibility was also assessed for back squat and bench press exercises done on an inertia-based device (16). As in the current study, repetitions were done against 4 submaximal loads for 2 workouts spaced 1 week apart (16). Yet unlike our study, each exercise only entailed concentric actions and was done on a machine that permitted barbell displacement along a linear motion arc. Interworkout CV values varied from 3.3 to 14.4. Our force-based indices fall within that same range and concur with prior results. Yet our power and velocity indices evoked higher CV values than did prior (16,18-20,27) exercise trials. In addition to the variability from higher movement rates (30), reasons for this disparity include a longer range of motion and ability to move in 3 planes of motion concurrently in this study.

Table 2 expresses SEM in both absolute and relative (SEM%) terms. An assessment tool of within-subject variability, the SEM is a common measure of absolute reliability (2,4) and was thus used in our study. Heteroscedasticity, or the greater dispersion to a series of paired scores as their absolute values increase, compromises SEM measurements, yet no evidence of heteroscedasticity exists for the current data (4). The SEM values in Table 2 are expressed in the same units of measurement as each performance index (11). Unlike SEM, our SEM% values assess the reproducibility of paired data without the impact of their absolute magnitude or value on results. The SEM and SEM% were used in nonexercise trials to assess data reproducibility (4,12). The SEM and SEM% values for gait performance tests, obtained from hemiparetic stroke victims, ranged from 0.07 to 18.6 and 4.8 to 8.2%, respectively (12). Data reproducibility was also assessed and obtained from the same types of subjects to note the impact illness had on their daily activities (4). The results showed that SEM% values that ranged from 3.34 to 14.53% (4). All but current P55 and V55 SEM% values compare favorably with the reproducibility values from stroke victims (4,12).

Recent exercise studies also used SEM and SEM% to assess data reproducibility (7-9). High-speed workouts done on exercise machines by healthy subjects yielded a wide range of SEM and SEM% values (8,9). The SEM varied from 6.2 to 430 and was dependent on the overall magnitude (score) of the variable under examination (8,9). Predictably, SEM% results included a smaller range (6.5–12.0) of values that were not impacted by their absolute score (9). The SEM values were also derived for various vertical jump indices done by men and women (7). Per subject, with multiple jumps performed at 2 exercise sessions spaced several days apart, SEM values were calculated for both intra and intertrial analyses (7). The SEM values ranged from 0.06 to 10.23 and 0.12 to 12.94 for the intratrial and intertrial analyses, respectively (7). The SEM and SEM% in Table 2 are generally higher than those reported previously (7-9) and relate to the greater skill required to perform a free weight (barbell) front squat. Yet, given the uniqueness of our data collection conditions (athletes performed a free weight exercise to improve their power as they prepared for their competitive season, workouts spaced 1 week apart), most of our performance indices met or exceeded a level of data reproducibility previously deemed as acceptable.

Back to Top | Article Outline

Practical Applications

Because of the data collection conditions of this study, our results are quite applicable to the workout conditions that athletes encounter as they prepare for their competitive season. Based on the current subjects and the location of data collection, we believe that our results may be more pertinent to coaches than prior (6-9,11,12,16,18-20,23,27) reproducibility studies are. Although current workouts were done as part of a mesocycle geared toward greater power development, perhaps the most difficult current study aspect to reproduce over successive workouts was the high movement rates at the lightest barbell loads, as evidenced by the P55 and V55 values. Yet, most of our performance indices either met or exceeded acceptable levels of data reproducibility. Thus, coaches who monitor the resistive exercise performance or progress of athletes with accelerometry over time should understand that values collected at lighter barbell loads inherently yield less data reproducibility and should thus pay attention to exercise repetition movement rates.

Back to Top | Article Outline


We wish to thank our subjects and The University of Tulsa's Athletic Department, for their involvement. J.F. Caruso, N.M. Olson, S.T. Taylor, J.R. McLagan, C.M. Shepherd, J.A. Borgsmiller, M.L. Mason, and R.D. Riner were participants in the Tulsa Undergraduate Research Challenge at The University of Tulsa.

Back to Top | Article Outline


1. Aragon-Vargas, LF. Evaluation of four vertical jump tests: Methodology, reliability, validity, and accuracy. Meas Phys Educ Exerc Sci 4: 215–228, 2000.
2. Atkinson, G and Nevill, AM. Statistical methods for assessing measurement error (reliability) in variables relevant to sports medicine. Sports Med 26: 217–238, 1998.
3. Bartko, JJ. The intraclass correlation coefficient as a measure of reliability. Psychol Rep 19: 3–11, 1966.
4. Beckerman, H, Roebroeck, ME, Lankhorst, GJ, Becher, JG, Bezemer, PD, and Verbeek, ALM. Smallest real difference, a link between reproducibility and responsiveness. Qual Life Res 10: 571–578, 2001.
5. Bemben, MG and McCalip, GA. Strength and power relationships as a function of age. J Strength Cond Res 13: 330–338, 1999.
6. Braidot, AA, Brusa, MH, Lestussi, FE, and Parera, GP. Biomechanics of front and back squat exercises. In: 16th Argentine Bioengineering Congress and the 5th Conference of Clinical Engineering, J Phy: Conf Ser 90. 01 2009. 2007.
7. Caruso, JF, Daily, JS, Olson, NM, McLagan, JR, Shepherd, CM, Drummond, JL, Walker, RH, and West, JO. Vertical jump data reproducibility from an instrumented platform. Isokinet Exerc Sci 2011, in press.
8. Caruso, JF, Hari, P, Coday, MA, Leeper, A, Ramey, ES, Monda, JK, Hastings, LP, and Davison, S. Performance evaluation of a high-speed inertial exercise trainer. J Strength Cond Res 22: 1760–1768, 2008.
9. Caruso, JF, Kucera, S, Hari, P, McLagan, JR, Olson, NM, Shepherd, CM, and Marshall, MR. Data reproducibility from an inertial kinetic exercise machine. J Strength Cond Res 24: 3081–3087, 2010.
10. Colley, R, Gorber, SC, and Tremblay, MS. Quality control and data reduction procedures for accelerometry-derived measures of physical activity. Health Rep 21: 1–7, 2010.
11. Eliasziw, M, Young, SL, Woodbury, MG, and Fryday-Field, K. Statistical methodology for the concurrent assessment of interrater and intrarater reliability: Using goniometric measurements as an example. Phys Ther 74: 777–788, 1994.
12. Flansbjer, U-B, Holmback, AM, Downham, D, Patten, C, and Lexall, J. Reliability of gait performance tests in men and women with hemiparesis after stroke. J Rehabil Med 37: 75–82, 2005.
13. Gullett, JC, Tillman, MD, Gutierrez, GM, and Chow, JW. A biomechanical comparison of back and front squats in healthy trained individuals. J Strength Cond Res 23: 284–292, 2008.
14. Hopkins, WG. Measures of reliability in sports medicine and science. Sports Med 30: 1–15, 2000.
15. Hopkins, WG, Hawley, JA, and Burke, LM. Design and analysis of research on sport performance enhancement. Med Sci Sports Exerc 31: 472–485, 1999.
16. Jidovtseff, B, Croisier, JL, Lhermerout, C, Serre, L, Sac, D, and Crielaard, JM. The concept of iso-inertial assessment: Reproducibility analysis and descriptive data. Isokinet Exerc Sci 4: 53–62, 2006.
17. Keppel, G, Saufley, WH, and Tokunaga, H. Introduction into Design and Analysis (2nd ed.). New York, NY: WH Freeman and Company, 1992.
18. Koshida, S, Iwai, K, Kagimori, A, and Urabe, Y. Contribution of maximal strength to peak power and rate of power development in bench press movement using free weights. Int Soc Biomech Sports 446–449, 2008.
19. Kovaleski, JE, Heitman, RJ, Gurchiek, LR, Erdmann, JW, and Trundle, TL. Reliability and effects of leg dominance on lower extremity isokinetic force and work using the closed chain rider system. J Sport Rehab 6: 319–326, 1997.
20. Kovaleski, JE, Ingersoll, CD, Knight, KL, and Mahar, CP. Reliability of the BTE dynatrack isotonic dynamometer. Isokinet Exerc Sci 6: 41–43, 1996.
21. Mayhew, JL, Piper, FC, and Ware, JS. Anthropometric correlates with strength performance among resistance trained athletes. J Sports Med Phys Fitness 33: 159–165, 1993.
22. Ottenbacher, KJ and Tomchek, SD. Measurement variation in method comparison studies: An empirical examination. Arch Phys Med Rehabil 75: 505–512, 1994.
23. Rontu, J-P, Hannula, MI, Leskinen, S, Linnamo, V, and Salmi, JA. One-repetition maximum bench press performance estimated with a new accelerometer method. J Strength Cond Res 24: 2018–2025, 2010.
24. Shrout, PE and Fleiss, JL. Intraclass correlation: Uses in assessing rater reliability. Psychol Bull 86: 420–428, 1979.
25. SMB380. Tri-axial acceleration sensor data sheet (preliminary). Bosch Sensortec (Rev. 1.2). July 16th 2007.
26. Stokes, M. Reliability and repeatability of methods for measuring muscle in physiotherapy. Physiother Pract 1: 71–76, 1985.
27. Thompson, CL and Bemben, MG. Reliability and comparability of the accelerometer as a measure of muscular power. Med Sci Sports Exerc 31: 897–902, 1999.
28. Trappe, SW, Trappe, TA, Lee, GA, and Costill, DL. Calf muscle strength in humans. Int J Sports Med 22: 186–191, 2001.
29. Wyse, JP, Mercer, TH, and Gleeson, NP. Time-of-day dependence of isokinetic leg strength and associated interday variability. Br J Sports Med 28: 167–170, 1994.
30. Zehr, EP and Sale, DG. Reproducibility of a ballistic movement. Med Sci Sports Exerc 29: 1383–1388, 1997.

deep knee bend; test-retest measures; data variability

Copyright © 2012 by the National Strength & Conditioning Association.