See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/374998761

Statistical Tests for Sports Science Practitioners: Identifying Performance

Gains in Individual Athletes

Article in The Journal of Strength and Conditioning Research · October 2023

DOI: 10.1519/JSC.0000000000004727

CITATIONS READS

2 2,994

4 authors, including:

John R Harry Chris Bishop

Texas Tech University Middlesex University, UK
91 PUBLICATIONS 1,208 CITATIONS 280 PUBLICATIONS 5,824 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by John R Harry on 17 November 2023.

The user has requested enhancement of the downloaded file.

 Statistical Tests for Sports Science Practitioners:
Identifying Performance Gains in Individual Athletes

John R. Harry, PhD1, Jacob Hurwitz, MS2, Connor Agnew, MS3, Chris Bishop, PhD4

1
Department of Kinesiology & Sport Management, Texas Tech University, Lubbock, TX

USA
2
Department of Kinesiology, Mississippi State University, Starkville, MS USA
3
Department of Athletics, Appalachian State University, Boone, NC USA
4
Faculty of Science and Technology, London Sport Institute, Middlesex University, London,

UK

Funding: This research did not receive any specific grant from funding agencies in the

public, commercial or not-for-profit sectors.

Conflict of Interest Disclosure: The authors have no conflicts of interest to disclose.

Correspondence Address: John R. Harry, PhD, CSCS; Dept. of Kinesiology & Sport

Management, Texas Tech University, 3204 Main Street, Lubbock, TX 79409 USA; Email:

john.harry@ttu.edu; Tel: (806) 834-7836; Twitter: @johnharry76

Running Title: Statistics for Individual Athlete Adaptations

Abstract

There is an ongoing surge of sports science professionals within sports organizations.

However, when seeking to determine training-related adaptations, sports scientists have

demonstrated continued reliance on group-style statistical analyses that are held to critical

assumptions not achievable in smaller-sample team settings. There is justification that these

team settings are better suited for replicated single-subject analyses, but there is a dearth of

literature to guide sports science professionals seeking methods appropriate for their teams.

In this report, we summarize four methods’ ability to detect performance adaptations at the

replicated single-subject level and provide our assessment for the ideal methods. These

methods included the model statistic, smallest worthwhile change (SWC), coefficient of

variation (CV), and standard error of measurement (SEM), which were discussed alongside

step-by-step guides for how to conduct each test. To contextualize the methods’ use in

practice, real countermovement vertical jump (CMJ) test data was used from four athletes

(two females and two males) who complete five bi-weekly CMJ test sessions. Each athlete

was competing in basketball at the NCAA Division 1 level. We concluded the combined

application of the model statistic and CV methods should be preferred when seeking to

objectively detect meaningful training adaptations in individual athletes. This combined

approach ensures that the differences between tests are A) not random and B) reflect a

worthwhile change. Ultimately, the use of simple and effective methods that are not restricted

by group-based statistical assumptions can aid practitioners when conducting performance

tests to determine athlete adaptations.

Keywords: data analysis; sports science; statistics

Introduction

Within sports organizations, there has been a concerted effort to form collaborative teams

capable of evaluating the effectiveness of sports performance interventions using objective,

scientific analysis methods. Typically, the main objective of this effort is to provide empirical

assessments of a sports program’s effectiveness by combining objective data with coaches’

expertise to help guide actionable decisions. This has led to the development of the field

known as “sports science” (16). Despite the benefits realized through this effort and the

development of the larger field of sports science, there has been little advancement with

respect to practices centering on individual athlete evaluations in place of team (i.e., group)

average evaluations. Ultimately, practitioners and team-based sport scientists will continue to

face challenges when seeking to determine whether athletes are adapting to training or related

interventions. At face value, these challenges are due to the available literature to which

practitioners can compare their athletes’ physical performances, as studies are largely limited

to short-term, repeated-measures and cross-sectional protocols that provide only mechanistic

explanations of physical human performance, rely on samples of non-athlete populations, or

both (1, 28, 32, 44, 59). In reality, multiple factors are to blame for the lack of ability to make

actionable decisions from collected data during performance tests.

Our experiences indicate the limited potential to make actionable decisions is predominantly

due to three realities within the strength and conditioning community. First, in the United

States there have been very few vacancies within sports organizations that are appealing to

qualified sports scientists, which forces untrained or inexperienced practitioners and scientists

into such roles. Second, strength and conditioning coaches are often assigned to, or volunteer

for, sports scientist roles despite a lack of training in areas concerned with research

methodology, data management and analysis, and statistics. Third, formally-trained scientists
who assume roles in sport are predominantly exposed to statistical tests aimed at generalizing

a sample’s average result to the sample’s larger population (26). Obviously, we cannot

control the current state of the literature, personnel decisions, nor the availability of appealing

vacancies for adequately trained candidates in sports organizations. However, we can provide

evidence- and experience-based methodologies we have used or continue to use in real-world

settings to help scientists and practitioners conduct appropriate analyses and obtain actionable

test results.

Conventional statistical analyses limit practitioners to the assessment of a group’s “average”

response or adaptation, and this practice is echoed in the sports science literature (4, 17, 18,

34, 50). However, replicated single-subject approaches have been used by some to detect

changes and this approach may provide the most value to sports scientists and practitioners

(11, 29, 31, 55), particularly high value for those working with smaller squads of athletes

(e.g., basketball, soccer, volleyball, etc.). An appropriate foundation of single-subject

analyses in athletes is forming (11, 29, 31), helping to create the impetus needed to move

away from group-average assessments, when necessary, without reliance on subjective visual

inspection or trend analyses. Unfortunately, anecdotal evidence indicates many practitioners

currently employing replicated single-subject assessments rely on arbitrary thresholds for

change (i.e., 5% or 10% improvement) as an indicator for real change. The next logical step

to strengthen the foundation is to demonstrate methods that can be used to conduct high-

quality objective assessments and provide empirical evidence for individual athletes’

performance adaptations. The ultimate objective from the current article was intended to

align with the NSCA’s Essentials of Sport Science textbook, which highlights the importance

of guiding necessary stakeholders in ways to quantitatively determine the effectiveness of

their training or related interventions and subsequently make data-driven changes (14).
The purpose of this report was to summarize methods for practitioners to explore for their

own purposes related to individual athlete assessments. We provide explanations for specific

methods’ selection compared to other options in addition to step-by-step guides to conduct

each test for further exploration. In addition, we compare each method and provide our

opinion for which method or combination of methods is most appropriate or ecologically

relevant for practitioners working with athlete populations. To contextualize how each

method can work in practice, we applied them to a small subset (n = 4) of real, longitudinal

data obtained from athletes competing in men’s or women’s basketball at the NCAA Division

1 level. Finally, we provide an editable Microsoft Excel worksheet which includes the

calculations to assist practitioners in applying our recommended approach in practice.

Motivation for Quantitatively Assessing Individual Athletes

It is beyond the scope of this report to discuss in-depth the limitations of group-level

statistical testing (e.g., t-test, ANOVA, etc.) in team-based settings, as this has been done

elsewhere (6, 8, 9, 22, 33). The key point for practitioners is that group-level testing requires

a normal distribution where the data from the sample (i.e., the athletes) are generalized prior

to analysis. As athlete samples are typically quite small, it is unlikely that a normal

distribution will occur, leading to issues related to the sample standard deviation and standard

error (13). Moreover, group-level approaches with smaller-samples are limited to the

interpretations based on the sample’s “average” response or adaptation, which is a well-

known misrepresentation of the individuals from which the group average is obtained (8).

Replicated single-subject approaches (11, 29, 31, 55) are ideal for smaller-sample settings

where the individuals are not represented by the group average. However, further examples

are needed to help practitioners move away from isolated use of group-average assessments,
when necessary, without reliance on subjective visual inspection or trend analyses for

individual assessments. Consequently, the next logical step is to demonstrate methods that

can be used to conduct high-quality assessments and provide empirical evidence for

individual athletes’ performance adaptations.

Importantly, there are multiple methods at one’s disposal to select from that can be used to

quantitatively explore individual athlete responses. However, some of those methods might

not be ideal even though they can be used at the individual level. Some of these methods,

albeit less common in contemporary literature, include non-parametric assessments such as

the Mann-Whitney U Test (9), bootstrapping (25), and multiple regression (24). Other

methods more commonly observed in contemporary literature include assessing the

percentage of non-overlapping data points (38), counting the number of data points above a

specific threshold (41), confidence intervals or effect sizes (52, 53), and statistical process

control (56). An advantage to these latter approaches is they provide adequate scientific rigor

because means and standard deviations of increasing or decreasing sequential data are treated

in consideration for their slope. This is important when comparisons are made between very

different means values (e.g., 10 vs. 100) with similar amounts of normalized variation (i.e.,

10%). The methods have additional value when seeking to determine whether an athletes’

current test result is different from a previous series of tests, or whether the results of one

training period are different from those of a subsequent training period. However, there are

consequences of these methods. First, they can involve between-subjects metrics, notably the

between-subject standard deviation, which minimizes the fluctuations or variability among

individuals within the team or group (8). Second, calculating rolling averages across several

test sessions minimized the movement variations (i.e., strategies) that dictate an athlete’s

performance outcomes during a test session when the test sessions are pooled together. As
described by Bates (6), all measurement outcomes are dependent upon the state of the

organism interacting with the environment at a specified moment in time. It is our opinion

that individual athlete assessments must seek to account for the uniqueness of individual

athlete strategy, as the available number of strategies changes over time due ongoing changes

in biomechanical or mechanical, morphological, and environmental constraints (6, 35). We

feel this is best accomplished by evaluating athletes’ performance changes between

consecutive test sessions. Those who may be interested in procedures that involve between-

subject metrics or rolling averages are referred to previous literature (36, 57, 64).

The quantitative approaches we feel have the most potential to reveal test-to-test performance

changes include single-subject ‘significance’ testing [i.e., the model statistic], and comparing

the magnitude of change against the smallest worthwhile change [SWC], the standard error of

measurement [SEM], or the coefficient of variation [CV] (10, 29, 31, 61, 65, 66). The

usefulness of each of these methods is that all were designed for, or can be relatively simple

to apply in, single-subject analyses. In addition, all are reasonably simple to understand,

perform, and interpret by practitioners, regardless of prior statistical training. Most

importantly, each method requires multiple testing efforts (i.e., trials) to be included for each

comparison, which discourages the common practice of including only the “best” effort or

only very few trials (42) from the test sessions. Including multiple trials is a critical

component of the single-subject methodology (6, 33) to reduce the variation in an athlete’s

data, thereby increasing statistical power (7) and provide stable performance data which

better reflects an athlete’s true performance capabilities (40). The resulting outcome for the

presence of a change is therefore obtained with consideration for both the absolute change of

performance and the potential variations among trials within the test (i.e., consistency of each

individual’s result), which can mask changes when not accounted for (8). Importantly,
accounting for variation means the performance result for each athlete’s test session(s)

considers their biomechanical constraints and unique response patterns to environmental

feedback, which ultimately, determine bodily movement and the amount of performance

variability (5, 35).

Overview of Session vs. Session Analysis Methods

Model Statistic

The model statistic technique is a critical difference method that can be loosely considered a

single-subject dependent t-test, whereby the observed difference between sessions is

compared to a probabilistic critical difference (7, 9). It was designed in the early 1990s by

Bates et al. (7) and has been used to demonstrate the value of single-subject comparisons

between two conditions relative to the group-level equivalent (31). Critical values were

generated (7) for selected trial sizes (i.e., the number of trials used to calculate the test session

average) and statistical probabilities (i.e., alpha levels; α), which are provided in Table 1. The

final decision from the test therefore indicates the probability for whether the difference

between test sessions was due to random chance, using the user’s a priori choice among

10%, 5%, or 1% probability levels. A unique feature of the model statistic is that it does not

calculate interval limits as one standard deviation away from the mean, and instead

incorporates the weighted mean standard deviation (7), which can also be described as the

variation in the collective number of trials or observations used in the comparison (7, 21).

Ultimately, the critical values and ultimately the critical difference score are analogous to a

1.96*standard deviation interval where the critical difference is a cutoff from which 95

percent of scores are between ± the critical difference. When test sessions with different trials

sizes are compared, we recommended using the critical value associated with the smallest

trial size because the test is more conservative when smaller trial sizes are used (i.e., more
difficult to return a difference that is not due to chance). The main limitation to the model

statistic is that the critical value table was created using vertical ground reaction force data

obtained during the support phase of running. It may be that other types of data or other tasks

might yield different critical values.

Traditional paired-samples t-test and effect sizes could be used in a similar way, but we favor

the model statistic for a few reasons. First, t-tests are inappropriate for repeated test

assessments (e.g., more than two comparative tests) without corrections for familywise error,

require a minimum number of samples (i.e., trials in this case) for adequate statistical power

(7), and the observations that make up the comparative means are subject assumptions related

to frequency distributions and homogeneity of variance (63). Satisfying these assumptions

and related criteria is not typically achievable in athlete testing environments. For effect

sizes, clinical data suggests a similar rate of “differences” will occur in comparison to the

model statistic (23). There are lingering problems, however, related to effect sizes. First, there

is no similar evidence suggesting a similar rate of difference among athlete populations

compared to the model statistic. Second, effect sizes are limited by the need to select a

magnitude threshold for interpretation based on subjective scales (19, 37, 54), created mostly

from recreationally active samples, or based on athlete training level and not primary sport or

training stimulus (54). As reducing subjectivity and guesswork is critical for the applied

sports scientist, we feel effect sizes would not be appropriate in this type of scenario.

Comparisons between tests for an individual athlete can be performed with the model statistic

technique using the following procedural steps:

1. Calculate the absolute mean difference between pre-test (X1) and post-test (X2)

sessions, ensuring a minimum of three trials for each test

Mean Difference = |X1 – X2|

 2. Calculate the mean standard deviation (SD) from the pre-test and post-test values

(SD1 and SD2)

SDMean = [(SD12 + SD22) / 2]1/2

3. Multiply desired critical value (i.e., test statistic) from Table 1 by the mean SD

calculated in step 2, which produces a critical difference

Critical Difference = Table 1 value * SDMean

4. Compare the mean difference between tests to the critical difference

- Significant Difference: Mean Difference > Critical Difference

- Non-Significant Difference: Mean Difference ≤ Critical Difference

< Insert Table 1 About Here >

Smallest Worthwhile Change (SWC)

The SWC approach is a form of magnitude-based inference (64) that was pioneered as an

attempt to provide a more realistic assessment of performance adaptations when compared to

traditional statistical tests (37). The SWC has greater specificity to sports science data sets

compared to conventional Fisher-based probability tests, which, as mentioned, tend to include

small sample (or trial) sizes, have relatively higher variations among measurements, and

sports scientists may be less concerned with whether a change is or is not due to

unexplainable chance (15, 47). A constant of 0.2 is used to establish the SWC threshold (see

below) for trained populations or athletes (60). This is because it aligns with the commonly

accepted, albeit subjective, definition for a “small” magnitude difference, or effect size (19).

A constant of 0.6 can be used in the SWC calculation for untrained populations or youth

athletes because large adaptations can be realized in those populations following initial or

short-term periods of training (47). While the selection of the SWC constant could be
objectively calculated (46) for a specific sample or athlete, we elected to use what we

determined from the sports science literature to be the most common SWC approach.

Although the SWC approach is typically utilized at the group level, it can be easily applied at

the replicated single-subject level. Comparisons between tests for an individual athlete can

performed with the SWC approach using the following procedural steps, which are slightly

modified for the single-subject assessment:

1. Calculate the athlete’s mean performance display for the pre-test (X1) and post-test

(X2) sessions

2. Calculate the change from X1 to X2

3. Calculate the SD across all test sessions (pre- and post-test sessions; referred to here

as SDGlobal)

4. Calculate the SWC

SWC = 0.2 * SDGlobal

5. Compare the athlete mean difference to the SWC

- True Difference: X2 > X1 + SWC

- Trivial Difference: X2 ≤ X1 + SWC

Coefficient of Variation (CV)

When comparing the percent change between test sessions to the CV, the objective is to

determine whether performance changes or differences between conditions are greater than

the variation in the test (10). Therefore, this technique is not a probabilistic test, and instead

reveals whether the observed difference exceeds the ‘noise’ (as represented by the CV)

inherent in the results. Rather than using the mean difference between the pre- and post-tests,

the percent difference is typically used and compared to the CV. This may be particularly

beneficial because the CV is a standardized metric that is a measure of reliability, which

enables the formation of a target or, if desired, a boundary used to detect positive and

negative change (60, 62). A benefit of this is that it may be more feasible to explain to

athletes, coaches, or other stakeholders “how” differences are determined for each

comparison. In addition, the CV approach can be obtained such that the unit of measure for

the test is retained and still tell the same story as when converted to a percentage value, which

can simplify interpretation for some practitioners, stakeholders, or both. While the procedure

outlined here uses one standard deviation to calculate the CV, 1.5 or 2 standard deviations

could also be used to expand the “range of scores”, which could be useful when seeking to

modify the sensitivity of the test and account for what is quantified by each metric’s interval.

This should not be confused with similar processes to compare data from different numeric

scales (64). Rather, it is a way to control the outcome sensitivity for performance tests known

to have greater or lesser movement variability. Comparisons between tests for an individual

athlete can be performed with the CV technique using the following procedural steps:

1. Calculate the athlete’s mean performance display for the pre-test (X1) and post-test

(X2) sessions

2. Calculate the athlete’s SD for the pre-test (SD1) session

3. Calculate the percent change from X1 to X2

% Change = 100 * (X2 – X1) / X1

4. Calculate the CV for the pre-test (CV1) session

CV1 = 100% * SD1 / X1

5. Compare the percent change between pre-test and post-test sessions to CV1

- True Difference: % Change > CV1

- Trivial Difference: % Change ≤ CV1

Standard Error of the Measurement (SEM)

For the SEM method, the objective is to compare the change in performance to the ‘noise’,

like the CV method, with data inherently retained in the original unit of measurement. This

means that the absolute mean difference (i.e., change of performance) is used in conjunction

with the precision of the test data. The SEM has been described as the intra-individual

version of the SD (13), but there appears to be lesser practical application of SEM because

assessments of individual performance changes have historically incorporated the SD (31).

Although there are two common formulae used to calculate the SEM (2), the inability to

obtain an intra-class correlation coefficient (ICC) at the single-subject level requires that the

SEM is calculated as the square root of the mean square error (MSE), with some

modifications to align with the single-subject dataset. This process may be more complicated

to some than using the ICC. However, it could be beneficial because it avoids the

uncertainties connected to the ICC and allows for more consistency (65) from test to test.

Procedural steps for using the SEM method are as follows:

1. Calculate the athlete’s mean performance display for the current test (X1) and the

mean of the two test sessions being compared (XTotal)

2. Calculate the sum of squares (SS)

SS = (X1 –XTotal)2

3. Calculate the MSE using the SS and degrees of freedom (df; number of trials -1),

where number of trials equals the sum of the number of trials recorded across sessions

used for step 1

MSE = SS/df

4. Calculate the SEM

SEM = √MSE

5. Compare the athlete mean difference to the SEM

 - True Difference: X2 > X1 + SEM

- Trivial Difference: X2 ≤ X1 + SEM

Summary of the Exemplar Athletes, Performance Task, and Metrics Used

To demonstrate the similarities and/or differences among methods with respect to detecting

performance changes in athletes, we used data obtained across five bi-weekly test sessions

from two female and two male NCAA Division 1 Basketball players. All were healthy,

uninjured, and active members of an NCAA Division 1 Basketball program throughout the

data collection period. Participants provided written informed consent, and data were

collected in accordance with the Declaration of Helsinki as approved by the local Institutional

Review Board. Although this report does not involve formal research methodology nor is it

an “original research” paper, it is a series of case examples using real data. Because of this,

we felt compelled to acknowledge the ethical considerations for using real data in this report.

The countermovement vertical jump (CMJ) was selected as the test activity, with ground

reaction force data obtained during testing. The CMJ was used for this report because it is

commonly used in research when seeking to understand physical ability among athlete

populations (3, 20, 28, 44, 49, 58). The CMJ was also selected because it (and related jumps)

is performed frequently during competitive play in basketball (27, 48) and strongly associated

with sport-specific qualities such as speed, strength, and agility (4, 45, 49, 51). In addition,

CMJ tests are routinely performed in laboratory and practitioner settings where multiple trials

are collected for each athlete, thereby satisfying the requirements of each method (see

below). The modified reactive strength index (RSIMOD) and vertical jump height were

included as primary and secondary CMJ performance metrics, calculated as center of mass

flight height and the ratio of vertical jump height and time to takeoff, respectively (30). These
metrics were selected according to a recent framework produced to guide practitioners in the

selection of useful metrics to examine CMJ abilities (12). This is because RSIMOD is a valid

and reliable surrogate for athletic explosiveness (43), and RSIMOD appears influenced

primarily by jump height and not time to takeoff (32).

Comparison among Methods at Detecting “Change”

Performance changes from session to session were determined using the model statistic,

SWC, CV, and SEM methods described previously. Table 2 provides a summary of the

cumulative increases of performance detected by each method (i.e., 4 possible changes per

method of analysis, per athlete, resulting in 16 possible changes). When detecting increases in

RSIMOD, the most sensitive methods were the SWC and CV methods, with both detecting an

increase of performance between test sessions for 38% (6 out of 16) of the total possible

comparisons (Table 2). The model statistic and SEM methods were more conservative,

detecting increases in performance between test sessions for 25% (4 out of 16) and 19% (3

out of 16) of the total possible comparisons, respectively (Table 2). For jump height, the most

sensitive method was SWC, with increases detected during 50% (8 out of 16) of the total

number of comparisons (Table 2). The next most sensitive method was the CV method,

detecting increases in jump height for 44% (7 out of 16) of the total number of comparisons

(Table 2). The model statistic method detected increases in jump height for 31% (5 out of 16)

of the number of comparisons, while the SEM method detected increases for only 6% (1 out

of 16) of the total number of comparisons (Table 2).

Importantly, the four methods were largely inconsistent with respect to detecting performance

increases for the same comparisons, as the methods were consistent during only ~13% (4 out
of 32) of the total possible comparisons across the four athletes (see Figures 1-4). The reason

the SWC detected a much greater number of performance gains versus all other methods is

that the equation uses only a portion of the athlete’s variation (20% for the 0.2 constant; 60%

for the 0.6 constant). This creates a scenario in which the SWC will inherently detect a

greater number of performance changes than the other methods, and ultimately, the risk of

false “gains” is greatest. As such, the methods should not be used interchangeably, nor should

reports of change be compared between or among assessments with different methods for

detecting change.

< Insert Table 2 About Here >

Critical Considerations for Detecting Actionable Change

The risk of false-positive outcomes should come into play from the perspective of

determining an actionable change. It is therefore our opinion that a test for change in team

settings includes three key components. First, the test should be objective to eliminate

subjective estimations or guesses. Second, it should it err on the side of conservativeness

when determining whether the difference between two tests is random or likely to be

legitimate to avoid erroneous interpretations. Third, the test should provide a simple way to

determine whether a difference between two tests is meaningful to the athlete, practitioner, or

related stakeholder. According to these requisite components and the benefits, limitations, or

both discussed for each method, we recommend the model statistic and CV approaches be

used in parallel. This combined use of approaches will provide practitioners with the ability

to identify actionable changes in their athletes. Further, it eliminates the likelihood for

decisions to be made based on the presence of potentially random differences that seem

valuable or non-random differences of little value. This “two-pronged” comparative approach

should be familiar to sports science practitioners and its value to practitioners and key

stakeholders can be reviewed elsewhere (37).

Table 3 shows the cumulative increases of performance based on the model statistic, CV, and

combined approach (the table is also available as an editable Microsoft Excel file, provided

with the article as supplemental digital content). The fact that the model statistic and CV

approaches did not identify athlete-specific differences with the same pattern supports the

potential use of the two methods in parallel. The reason for the different patterns of

differences between the approaches relates to the objectives for which the tests were

designed. From the sports scientist’s perspective, the model statistic indicates whether the test

results are legitimately different due to its conservativeness, while the CV indicates the

meaningfulness of the difference between the test results. As such, they are complementary

methods as opposed to ones that should be compared against one another.

< Insert Table 3 >

The recommendation for using the model statistic may be obvious, as it is the only method

providing whether the difference between tests is statistically significant like convention

group analyses (e.g., t-test, ANOVA). However, the recommendation for parallel use of the

CV method may be less clear because the CV, SEM, and SWC methods all incorporate the

session variation, or the athlete’s consistency across trials performed with the session(s). Our

position for omitting the SWC relates to its reliance on subjectivity when determining the

threshold for importance of a change. Moreover, the SWC’s use of 20% or 60% of the

athlete’s variation makes it excessively sensitive, as mentioned previously, which is not ideal

for athlete populations with a wide range of responses to training stimuli.

The main limitation to the SEM method is that it involves modified equations to permit its

use at the individual athlete level. In addition, the SEM method returned the same number of

differences as the SWC method, albeit with a similar yet not identical pattern of differences

(not shown in tables or figures). Thus, there may be increased risk for erroneous conclusions

using the SWC or SEM methods, in which a certain number of significant performance

changes from the model statistic would coincide with an inflated number of “meaningful”

changes versus the CV method. In turn, practitioners using SWC or SEM might be motivated

to conclude that an athlete has demonstrated a positive adaptation when there was no

adaptation, or the adaptation was trivial.

Practical Applications: When to Make Data-Driven Changes

One final point to note relates to the level of influence a practitioner or sports scientist places

on the results of a single or repeated performance tests, as the importance of any test can vary

among practitioners, athletes, or both depending on training objectives. For context, Figures

1-4 demonstrate our individual athletes’ CMJ performance changes across five sessions (i.e.,

10 weeks) from the start of full-time training until approximate start of the competition

season. The nature of the training interventions for each athlete, which was somewhat unique

to each individual, centered on progressive changes in training volume and intensity for

exercises to continue improving “explosive strength” (67). As mentioned, we operationally

defined “increased explosive strength” as an increase of RSIMOD, with jump height used as a

secondary performance metric to help explain changes in RSIMOD. We use this approach

because of previous work demonstrating RSIMOD is a valid and reliable surrogate for athletic

explosiveness (43) in addition to our work suggesting increased RSIMOD is influenced

primarily by jump height and not its other component part, time to takeoff (32). As such, if

jump height was not an adequate explanatory metric for RSIMOD changes, it would mean the
change was due primarily to altered times to takeoff and that would be considered in any

recommendations provided. Thus, all athletes were expected to realize CMJ performance

gains at each test session. If a change was not observed on a given test day, we would reflect

on the results and other contributing factors to overall workload, such as physical training,

on-court work, test-day fatigue/athlete readiness. Those data will not be discussed here but

can be explored in part elsewhere (39). According to those reflections, recommendations

were given to the strength and conditioning staff to decide whether training or related

changes were needed.

< Insert Figures 1-4 About Here >

For female athlete 1 (Figure 1), the session-to-session changes in RSIMOD were similar to

their changes in jump height. For the purposes of this report, we will focus on the way in

which we use CMJ results to make data-driven training changes. In particular, the large

decreases in RSIMOD and jump height from test session 2 to 3 were concerning. A member of

the sports science team provided the strength and conditioning staff with recommendations,

and they decided whether training and/or related workload modifications were appropriate.

The changes they implemented were shown to be successful according to the athlete’s CMJ

results at test session 4, which were both statistically significant (model statistic) and

meaningful (CV). For female athlete 2, the CMJ test results were not concerning enough to

recommend specific changes or considerations between test sessions 1 and 5, though there

was a statistically significant (model statistic) and meaningful (CV) increase of RSIMOD at

test session 5.

It is important to note that our training, recommendation-based changes, or both, do not

always return positive changes, as shown between test sessions 4 and 5 for male athlete 1
(Figure 3). For instance, this athlete demonstrated somewhat positive results from test session

1 to 2, as RSIMOD was shown to meaningfully increase (CV) while jump height was shown to

increase significantly (model statistic) and meaningfully (CV). This performance further

improved from test session 2 to 3, as RSIMOD and jump height increased significantly (model

statistic) and meaningfully (CV) increased. However, there was an alarming drop in

performance, as evidenced by decreases in both RSIMOD and jump height. Although data-

driven recommendations were presented and training and/or related modifications were

prescribed, the athlete did not display statistically significant (model statistic) nor meaningful

(CV) improvements in RSIMOD at test session 5 due to substantial variation across the session

5 trials. Further to this, male athlete 2 displayed concerning results from test session 1 to 2,

where a ~30% decrease in RSIMOD occurred. As the RSIMOD decrease appeared to be

primarily driven by time to takeoff due to the < 5% decrease in jump height, specific

recommendations and training and/or related modifications were prescribed specific to the

athlete’s display. Interestingly, the athlete’s subsequent results at test session 3 did not

stimulate targeted change in RSIMOD nor jump height. This trend continued through test

session 5, suggesting the athlete was, for some reason, more resistant to the changes we were

recommending, or the strength and conditioning staff were prescribing. While detailed

exploration into those reasons is beyond the scope of this report, it did provide insight into

ways test results could be used to consider or better adjust training.

A final point for this section relates to our objectives during the 10-week test period

presented, as this approach was specific to the groups of athletes and the goals established for

those athletes. While the overarching objective for this subset of athletes and most athletes in

the basketball programs is to increase “explosive strength”, the way in which that is

stimulated varies across seasons, training blocks, and athletes. This is why the specific
training programs, recommendations, and complimentary data are not discussed, as it would

be of little relevance to the reader. The objective for this practical application section was to

show when test data reveals performance changes that are objective in nature and should

stimulate some consideration as to whether a modification is needed for training or some

related aspect of the overall program.

Conclusion

We summarized four methods, specifically the model statistic, smallest worthwhile change

(SWC), coefficient of variation (CV), and standard error of the measurement (SEM), with

respect to detecting individual athletes’ change during performance tests, using the CMJ as

the example test. We provide support for our recommendation that the combined use of the

model statistic and CV should be preferred when seeking to objectively detect real and

important training adaptations in individual athletes. We applied our recommended approach

to a small subset of real data obtained in four different athletes, competing in men’s or

women’s basketball at the NCAA Division 1 level, to contextualize how these methods and

can work in practice, highlighting when we would or would not recommend or prescribe a

training-related modification.
 References

1. Adams K, O’Shea JP, O’Shea KL, and Climstein M. The effect of six weeks of squat,

plyometric and squat-plyometric training on power production. Journal of applied

sport science research 6: 36-41, 1992.

2. Atkinson G and Nevill AM. Statistical methods for assessing measurement error

(reliability) in variables relevant to sports medicine. Sports medicine 26: 217-238,

1998.

3. Barker LA, Harry JR, and Mercer JA. Relationships Between Countermovement

Jump Ground Reaction Forces and Jump Height, Reactive Strength Index, and Jump

Time. Journal of Strength & Conditioning Research 32(1): 248-254, 2018.

4. Barnes JL, Schilling BK, Falvo MJ, Weiss LW, Creasy AK, and Fry AC. Relationship

of jumping and agility performance in female volleyball athletes. Journal of strength

and conditioning research / National Strength & Conditioning Association 21: 1192-

1196, 2007.

5. Bates BT. Scientific basis of human movement. Journal of Physical Education &

Recreation 48: 70, 1977.

6. Bates BT. Single-subject methodology: an alternative approach. Medicine & Science

in Sports & Exercise 28: 631-638, 1996.

7. Bates BT, Dufek JS, and Davis HP. The effect of trial size on statistical power.

Medicine & Science in Sports & Exercise 24: 1059-1065, 1992.

8. Bates BT, Dufek JS, James CR, Harry JR, and Eggleston JD. The Influence of

Experimental Design on the Detection of Performance Differences. Measurement in

Physical Education and Exercise Science 20: 200-207, 2016.

9. Bates BT, James CR, and Dufek JS. Single subject analysis, in: Innovative Analyses

of Human Movement. N Stergiou, ed. Champaign, IL: Human Kinetics, 2004, pp 3-

28.

10. Bishop C, Abbott W, Brashill C, and Read P. Effects of Strength Training on Bilateral

and Unilateral Jump Performance, and the Bilateral Deficit in Premier League

Academy Soccer Players. UK Strength and Conditioning Association Published ahead

of print, 2021.

11. Bishop C, Lake J, Loturco I, Papadopoulos K, Turner AN, and Read P. Interlimb

asymmetries: The need for an individual approach to data analysis. Journal of

Strength and Conditioning Research 35: 695-701, 2021.

12. Bishop C, Turner A, Jordan M, Harry JR, Loturco I, Lake J, and Comfort P. A

Framework to Guide Practitioners when Selecting Metrics During the

Countermovement and Drop Jump Tests. Strength & Conditioning Journal In Press,

2021.

13. Bland JM and Altman DG. Measurement error. BMJ: British medical journal 312:

1654, 1996.

14. Brewer C. Performance Interventions and Operationalizing Data, in: NSCA's

Essentials of Sport Science. DN French, L Torres Ronda, eds. Champaign, IL: Human

Kinetics, 2022, pp 339-351.

15. Buchheit M. Chasing the 0.2. Int J Sports Physiol Perform 11: 417-418, 2016.

16. Burwitz L, Moore PM, and Wilkinson DM. Future directions for performance‐related

sports science research: An interdisciplinary approach. Journal of sports sciences 12:

93-109, 1994.
17. Castagna C and Castellini E. Vertical jump performance in Italian male and female

national team soccer players. Journal of strength and conditioning research / National

Strength & Conditioning Association 27: 1156-1161, 2013.

18. Cohen DD, Restrepo A, Richter C, Harry JR, Franchi MV, Restrepo C, R. P, and

Taberner M. Detraining of specific neuromuscular qualities in elite footballers during

COVID-19 quarantine. Science & Medicine in Football In Press, 2020.

19. Cohen J. A power primer. Psychological bulletin 112: 155, 1992.

20. Drinkwater EJ, Pyne DB, and McKenna MJ. Design and interpretation of

anthropometric and fitness testing of basketball players. Sports medicine 38: 565-578,

2008.

21. Dufek JS, Bates BT, Davis HP, and Malone LA. Dynamic performance assessment of

selected sport shoes on impact forces. Medicine and science in sports and exercise 23:

1062-1067, 1991.

22. Dufek JS, Bates BT, Stergiou N, and James CR. Interactive effects between group and

single-subject response patterns. Human Movement Science 14: 301-323, 1995.

23. Dufek JS, Harry JR, Eggleston JD, and Hickman RA. Walking mechanics and

movement pattern variability in monozygotic twins with autism Spectrum disorder.

Journal of Developmental and Physical Disabilities 30: 793-805, 2018.

24. Dufek JS and Zhang S. Landing models for volleyball players: a longitudinal

evaluation. The Journal of sports medicine and physical fitness 36: 35-42, 1996.

25. Edgington ES. Randomized single-subject experiments and statistical tests. Journal of

Counseling Psychology 34: 437, 1987.

26. Fisher RA. On the mathematical foundations of the theory of statistics, in: In: Theory

of Statistical Estimation (Proceedings of the Cambridge Philosophical Society).

England: Cambridge Philosophical Society, 1925, pp 700-725.

27. Hammond KG, Harry JR, Hays CW, and Schilling BK. Time-motion analysis of

men's professional beach volleyball. Journal of Sport and Human Performance 8,

2020.

28. Harry JR, Barker LA, James CR, and Dufek JS. Performance Differences Among

Skilled Soccer Players of Different Playing Positions During Vertical Jumping and

Landing. The Journal of Strength & Conditioning Research 32(2): 304-312, 2018.

29. Harry JR, Barker LA, Tinsley GM, Krzyszkowski J, Chowning L, McMahon JJ, and

Lake J. Relationships among countermovement vertical jump performance metrics,

strategy variables, and inter-limb asymmetry in females. Sports Biomechanics In

Press, 2021.

30. Harry JR, Blinch J, Barker LA, Krzyszkowski J, and Chowning L. Low pass filter

effects on metrics of countermovement vertical jump performance. The Journal of

Strength & Conditioning Research 36: 1459-1467, 2022.

31. Harry JR, Eggleston JD, Dufek JS, and James CR. Single-Subject Analyses Reveal

Altered Performance and Muscle Activation during Vertical Jumping. Biomechanics

1: 15-28, 2021.

32. Harry JR, Paquette MR, Schilling BK, Barker LA, James CR, and Dufek JS. Kinetic

and electromyographic sub-phase characteristics with relation to countermovement

vertical jump performance. Journal of Applied Biomechanics 34: 291-297, 2018.

33. Hecksteden A, Kraushaar J, Scharhag-Rosenberger F, Theisen D, Senn S, and Meyer

T. Individual response to exercise training-a statistical perspective. Journal of applied

physiology 118: 1450-1459, 2015.

34. Heishman A, Daub B, Miller R, Brown B, Freitas E, and Bemben M.

Countermovement jump inter-limb asymmetries in collegiate basketball players.

Sports 7: 103, 2019.

35. Higgins S. Movement as an emergent form: Its structural limits. Human Movement

Science 4(2): 119-148, 1985.

36. Hopkins WG and Batterham AM. Error rates, decisive outcomes and publication bias

with several inferential methods. Sports Medicine 46: 1563-1573, 2016.

37. Hopkins WG, Marshall SW, Batterham AM, and Hanin J. Progressive statistics for

studies in sports medicine and exercise science. Medicine & Science in Sports &

Exercise 41: 3, 2009.

38. Huberty CJ and Lowman LL. Group overlap as a basis for effect size. Educational

and Psychological Measurement 60: 543-563, 2000.

39. Hurwitz J, Rich D, Agnew C, and Harry J. Workload Management Strategies to

Optimize Athlete Performance in Collegiate Men’s and Women’s Basketball.

International Journal of Strength and Conditioning 2, 2022.

40. James CR, Herman JA, Dufek JS, and Bates BT. Number of trials necessary to

achieve performance stability of selected ground reaction force variables during

landing. Journal of sports science & medicine 6: 126, 2007.

41. Kazdin AE and Tuma AH. Single-case research designs. 1982.

42. Kennedy RA and Drake D. Improving the signal-to-noise ratio when monitoring

countermovement jump performance. The Journal of Strength & Conditioning

Research 35: 85-90, 2021.

43. Kipp K, Kiely MT, and Geiser CF. Reactive strength index modified is a valid

measure of explosiveness in collegiate female volleyball players. The Journal of

Strength & Conditioning Research 30: 1341-1347, 2016.

44. Krzyszkowski J, Chowning LD, and Harry JR. Phase-specific predictors of

countermovement jump performance that distinguish good from poor jumpers. The

Journal of Strength & Conditioning Research 36: 3352-3358, 2022.

45. Loturco I, Pereira LA, Cal Abad CC, D'Angelo RA, Fernandes V, Kitamura K, Kobal

R, and Nakamura FY. Vertical and Horizontal Jump Tests Are Strongly Associated

With Competitive Performance in 100-m Dash Events. Journal of strength and

conditioning research 29: 1966-1971, 2015.

46. Mann JB, Ivey PA, Mayhew JL, Schumacher RM, and Brechue WF. Relationship

between agility tests and short sprints: Reliability and smallest worthwhile difference

in National Collegiate Athletic Association Division-I football players. The Journal of

Strength & Conditioning Research 30: 893-900, 2016.

47. Marocolo M, Simim MAM, Bernardino A, Monteiro IR, Patterson SD, and da Mota

GR. Ischemic preconditioning and exercise performance: shedding light through

smallest worthwhile change. European journal of applied physiology 119: 2123-2149,

2019.

48. McInnes S, Carlson J, Jones C, and McKenna MJ. The physiological load imposed on

basketball players during competition. Journal of sports sciences 13: 387-397, 1995.

49. McMahon J, Lake JP, Ripley N, and Comfort P. Vertical jump testing in rugby

league: a rationale for calculating take-off momentum. Journal of Applied

Biomechanics In Press, 2020.

50. McMahon JJ, Murphy S, Rej SJ, and Comfort P. Countermovement-jump-phase

characteristics of senior and academy rugby league players. International journal of

sports physiology and performance 12: 803-811, 2017.

51. Nuzzo JL, McBride JM, Cormie P, and McCaulley GO. Relationship between

countermovement jump performance and multijoint isometric and dynamic tests of

strength. Journal of strength and conditioning research 22: 699-707, 2008.

52. Odgaard EC and Fowler RL. Confidence intervals for effect sizes: Compliance and

clinical significance in the Journal of Consulting and Clinical Psychology. Journal of

consulting and clinical psychology 78: 287, 2010.

53. Parker RI and Hagan-Burke S. Useful effect size interpretations for single case

research. Behavior therapy 38: 95-105, 2007.

54. Rhea MR. Determining the magnitude of treatment effects in strength training

research through the use of the effect size. Journal of strength and conditioning

research 18: 918-920, 2004.

55. Robertson S, Bartlett JD, and Gastin PB. Red, amber, or green? Athlete monitoring in

team sport: the need for decision-support systems. International journal of sports

physiology and performance 12: S2-73-S72-79, 2017.

56. Sands W, Cardinale M, McNeal J, Murray S, Sole C, Reed J, Apostolopoulos N, and

Stone M. Recommendations for measurement and management of an elite athlete.

Sports 7: 105, 2019.

57. Sands WA, Kavanaugh AA, Murray SR, McNeal JR, and Jemni M. Modern

techniques and technologies applied to training and performance monitoring.

International journal of sports physiology and performance 12: S2-63-S62-72, 2017.

58. Sporis G, Jukic I, Ostojic SM, and Milanovic D. Fitness profiling in soccer: physical

and physiologic characteristics of elite players. The Journal of Strength &

Conditioning Research 23: 1947-1953, 2009.

59. Stone MH, Sands WA, and Stone ME. The downfall of sports science in the United

States. Strength & Conditioning Journal 26: 72-75, 2004.

60. Turner A, Brazier J, Bishop C, Chavda S, Cree J, and Read P. Data analysis for

strength and conditioning coaches: Using excel to analyze reliability, differences, and

relationships. Strength & Conditioning Journal 37: 76-83, 2015.

61. Turner AN, Parmar N, Jovanovski A, and Hearne G. Assessing group-based changes

in high-performance sport. Part 1: null hypothesis significance testing and the utility

of p values. Strength & Conditioning Journal 43: 112-116, 2021.

62. Turner AN, Parmar N, and Jovonoski A. Assessing group-based changes in high-

performance sport. Part 2: Effect sizes and embracing uncertainty through confidence

intervals. Strength Cond J, 2021.

63. Vincent W and Weir J. Statistics in Kinesiology. Champaign, IL: Human Kinetics,

2012.

64. Ward P, Coutts AJ, Pruna R, and McCall A. Putting the “I” back in team.

International journal of sports physiology and performance 13: 1107-1111, 2018.

65. Weir JP. Quantifying test-retest reliability using the intraclass correlation coefficient

and the SEM. The Journal of Strength & Conditioning Research 19: 231-240, 2005.

66. Williams CA, Oliver JL, and Faulkner J. Seasonal monitoring of sprint and jump

performance in a soccer youth academy. International journal of sports physiology

and performance 6: 264-275, 2011.

67. Zatsiorsky VM, Kraemer WJ, and Fry AC. Task-Specific Strength, in: Science and

Practice of Strength Training. Champaign Illinois: Human Kinetics, 2021, p 25.

 Tables

Table 1. Critical values to determine the critical difference for the model statistic technique.

Trial Size α = 0.10 α = 0.05 α = 0.01

3 1.3733 1.6533 2.2133
4 1.2643 1.5058 1.9867
5 1.1597 1.3662 1.7788
6 1.0629 1.2408 1.6044
7 0.9751 1.1306 1.4623
8 0.8960 1.0351 1.3473
9 0.8270 0.9536 1.2542
10 0.7673 0.8857 1.1776
11 0.7172 0.8307 1.1129
12 0.6757 0.7867 1.0581
13 0.6415 0.7516 1.0117
14 0.6132 0.7234 0.9720
15 0.5896 0.7001 0.9375
16 0.5695 0.6798 0.9070
17 0.5522 0.6618 0.8796
18 0.5371 0.6458 0.8548
19 0.5237 0.6311 0.8318
20 0.5114 0.6175 0.8102
25 0.4592 0.5572 0.7145
30 0.4194 0.5097 0.6437
35 0.3896 0.4729 0.5949
40 0.3673 0.4442 0.5626
45 0.3500 0.4207 0.5414
50 0.3352 0.4000 0.5256

Notes – Trial size: the number of trials used to calculate the performance means for the test
sessions; α: the alpha level (i.e., statistical probability criterion), or the probability that a
difference between sessions is random.
Table 2. Performance increases detected by each method across athletes.

Model Statistic SWC CV% SEM

Metric Increases % Total Increases % Total Increases % Total Increases % Total
RSIMOD 4 25 8 50 6 38 8 50
Jump
5 31 10 63 6 38 10 63
Height

Notes – SWC: smallest worthwhile change method; CV% coefficient of variation method, SEM: standard error of measurement method;
Increases: number of increases detected across all four athletes (decreases excluded); % Total: percent of increases detected relative to the total
number of comparisons.
 Table 3. Performance improvement assessments across five test sessions and four athletes.

Test 1 Improvement Test 2 Improvement Test 3 Improvement Test 4 Improvement Test 5 Improvement

Athlete Mean SD MS CV Both Mean SD MS CV Both Mean SD MS CV Both Mean SD MS CV Both Mean SD MS CV Both

F1 0.393 0.016 N/A N/A N/A 0.547 0.043 TRUE TRUE TRUE 0.389 0.010 FALSE FALSE FALSE 0.560 0.030 TRUE TRUE TRUE 0.504 0.113 FALSE FALSE FALSE
RSIMOD
F2 0.337 0.007 N/A N/A N/A 0.367 0.051 FALSE TRUE FALSE 0.346 0.031 FALSE FALSE FALSE 0.327 0.010 FALSE FALSE FALSE 0.342 0.005 TRUE TRUE TRUE

M1 0.504 0.028 N/A N/A N/A 0.553 0.043 FALSE TRUE FALSE 0.688 0.058 TRUE TRUE TRUE 0.595 0.021 FALSE FALSE FALSE 0.532 0.160 FALSE FALSE FALSE

M2 0.777 0.126 N/A N/A N/A 0.536 0.012 FALSE FALSE FALSE 0.524 0.093 FALSE FALSE FALSE 0.526 0.076 FALSE FALSE FALSE 0.576 0.058 FALSE FALSE FALSE

Group 0.503 0.065 N/A N/A N/A 0.501 0.040 FALSE FALSE FALSE 0.487 0.057 FALSE FALSE FALSE 0.502 0.042 FALSE FALSE FALSE 0.489 0.102 FALSE FALSE FALSE

Test 1 Improvement Test 2 Improvement Test 3 Improvement Test 4 Improvement Test 5 Improvement

Athlete Mean SD MS CV Both Mean SD MS CV Both Mean SD MS CV Both Mean SD MS CV Both Mean SD MS CV Both

F1 0.248 0.001 N/A N/A N/A 0.306 0.012 TRUE TRUE TRUE 0.260 0.012 FALSE FALSE FALSE 0.301 0.012 TRUE TRUE TRUE 0.301 0.008 FALSE FALSE FALSE
Jump
Height F2 0.254 0.010 N/A N/A N/A 0.275 0.007 TRUE TRUE TRUE 0.263 0.013 FALSE FALSE FALSE 0.277 0.003 FALSE FALSE FALSE 0.265 0.012 FALSE FALSE FALSE

M1 0.319 0.010 N/A N/A N/A 0.339 0.009 TRUE TRUE TRUE 0.374 0.019 TRUE TRUE TRUE 0.355 0.003 FALSE FALSE FALSE 0.378 0.030 FALSE TRUE FALSE

M2 0.452 0.012 N/A N/A N/A 0.424 0.027 FALSE FALSE FALSE 0.405 0.029 FALSE FALSE FALSE 0.412 0.047 FALSE FALSE FALSE 0.438 0.008 FALSE FALSE FALSE

Group 0.318 0.009 N/A N/A N/A 0.336 0.016 FALSE TRUE FALSE 0.326 0.020 FALSE FALSE FALSE 0.336 0.024 FALSE FALSE FALSE 0.346 0.017 FALSE FALSE FALSE

Notes – F1: female athlete 1, F2: female athlete 2; M1: male athlete 1; M2: male athlete 2; MS: model statistic method; CV: coefficient of
variation method; Both: combined approach using model statistic and coefficient of variation methods; TRUE: change detected by associated
method; FALSE: change not detected by associated method; When TRUE is contained in green, both methods detected change.
 Figure Captions

Figure 1. RSIMOD (left) and jump height (right) performance results and differences detected by the model statistic and CV methods for female
athlete 1.

Notes – Data are presented as mean ± 1 standard deviation across 3 trials for each test session; Model Statistic Change from Previous: non-
random (p < 0.05) change between adjacent test sessions; CV% Minimum Increase: threshold that must be exceeded to indicate change between
adjacent test sessions.
Figure 2. RSIMOD (left) and jump height (right) performance results and differences detected by the model statistic and CV methods for female
athlete 2.

Notes – Data are presented as mean ± 1 standard deviation across 3 trials for each test session; Model Statistic Change from Previous: non-
random (p < 0.05) change between adjacent test sessions; CV% Minimum Increase: threshold that must be exceeded to indicate change between
adjacent test sessions.
Figure 3. RSIMOD (left) and jump height (right) performance results and differences detected by the model statistic and CV methods for male
athlete 1.

Notes – Data are presented as mean ± 1 standard deviation across 3 trials for each test session; Model Statistic Change from Previous: non-
random (p < 0.05) change between adjacent test sessions; CV% Minimum Increase: threshold that must be exceeded to indicate change between
adjacent test sessions.
 Figure 4. RSIMOD (left) and jump height (right) performance results and differences detected by the model statistic and CV methods for male
athlete 2.

Notes – Data are presented as mean ± 1 standard deviation across 3 trials for each test session; Model Statistic Change from Previous: non-
random (p < 0.05) change between adjacent test sessions; CV% Minimum Increase: threshold that must be exceeded to indicate change between
adjacent test sessions.

View publication stats