DISCRIMINANT ANALYSIS OF ANTHROPOMETRIC AND BIOMOTOR VARIABLES

AMONG ELITE ADOLESCENT FEMALE ATHLETES IN FOUR SPORTS.

Keywords: adolescents; elite athletes; female; anthropometric; biomotor.

1
ABSTRACT

The purpose of this study was to identify anthropometric and biomotor determinants that

discriminated among groups of elite adolescent female athletes (n=100, age=14.3 ± 1.3)

from four different sports (tennis n=15, swimming n=23, figure skating n=46 and

volleyball n=16). Anthropometric variables included body mass, height, biepicondilar

breadth of the distal extremity of the humerus and femur, maximal girth of the calf and

biceps and the sum of five adipose skinfolds. Biomotor variables were maximal aerobic

power, muscular endurance and flexibility of the trunk. Discriminant analysis revealed

three significant functions (P<0.05). The first discriminant function (DF1) primarily

represented differences between figure skaters and all other groups of athletes. Both

DF2 and DF3 underlined anthropometric and biomotor differences between swimmers

and volleyball players and between tennis players and swimmers respectively. After

validation, the analysis showed that 88% of the athletes were correctly classified in their

respective sports. Our model confirms that already at adolescence, elite female athletes

present physical and biomotor particularities that distinguish them clearly according to

their particular sport.

2
 INTRODUCTION

Over the past few years, particular attention has been focused on the description

of biomotor and anthropometric factors that distinguish elite athletes involved in different

sports (Orvana, 1987; Sovak and Hawes, 1987; Thomas et al., 1989; Smith and

Thomas, 1991; Gualdi-Russo et al., 1992; Loko et al., 2000). Results of this research

suggest that each sport is characterized by athletes having specific physical and

biomotor particularities favouring the performance in the given sport. If the identification

of variables that characterize athletes of different sports is of interest to understand what

distinguishes them, their relative contribution is also important. These aspects have

received little attention, particularly for female adolescent athletes who have undergone

intensive training (Housh et al., 1984b; Fogelholm et al., 2000).

Compared to adult athletes, who have already reached a high level of

performance, adolescents involved in the pursuit of athletic excellence are still in the

development phase (physically and technically) in many sports. Currently, there are few

references in the literature (Housh et al., 1984a) that allow to confirm as to whether or

not the description of both the anthropometric and the biomotor profiles is sufficiently

defined to characterize adolescent female athletes as a function of the sport practiced.

The purpose of this study was to develop a multivariate model that would allow

the identification and quantification of the individual contribution of a series of

anthropometric and biomotor variables to the discrimination of athletes according to their

sport.

METHODS

Subjects/Groups

One hundred french speaking Caucasian elite adolescent female athletes aged

between 12 and 17 years (mean age 14.3 ± 1.3 years) participated in this study. The

3
sample included subjects from four different sports; figure skating (n=46), swimming

(n=23), tennis (n=15) and volleyball (n=16). The investigation took place among

participants in a special scholastic program targeted to athletes from Montréal and the

surrounding area. All the subjects were classified among the best 15 athletes in the

province of Québec in their respective sport and age category. About 40% of the athletes

sampled competed at the national or international level. The period of competition was

spread out over a minimum of 10 consecutive months, which included an average of 18

hours/week of training in their respective sport. Written consent was obtained from the

parents for all subjects.

Measures

To assess the relative contribution of each set of parameters, the measured

variables were of two categories: 7 anthropometric and 4 biomotor variables. Most of

the selected variables were chosen from a battery of tests conceived for the project, and

came from the wide variety which have previously been used to describe athletic

populations (Carter, 1970; MacDougall et al., 1988; Carter and Ackland, 1994). For

consistency, all measurements were taken by the same tester.

All subjects were measured on the following anthropometric parameters: body

mass, height, biepicondilar breadth of the distal extremity of the humerus and femur,

maximal girth of the biceps and the calf, and the sum of five skinfold thicknesses

(triceps, abdomen, calf, suprailiac and subscapular). All measurements were taken

according to standardized procedures (Callaway et al., 1988; Wilmore et al., 1988).

The biomotor variables included maximal aerobic power (VO 2max) measured on

a treadmill (speed of 10 km/h with an increasing grade of 2.5% every 2.5 minutes until

exhaustion). The O2 and CO2 concentrations were measured by a breath-by-breath

metabolic analyzer (Sensormedics model 2900c, Anaheim, CA). Muscular endurance

was assessed by mean of two imposed rhythm tests by metronome; push-ups (50 push-

4
ups per minute till exhaustion) and burpees (25 burpees per minute till exhaustion)

(Leone and Léger, 1985). Finally, flexibility was evaluated by mean of the sit-and-reach

test (Fleishman, 1965).

STATISTICAL ANALYSIS

Multivariate analysis of variance (MANOVA) was first used to determine if the

groups were different. Discriminant analysis was then employed on 12 variables

measured, which included anthropometric and biomotor parameters, to develop a model

to predict membership of each athlete in the four different sports. Validation of the

discriminant model was conducted by the “jackknife” procedure (Norusis, 1993). Only

subjects with complete data sets were retained for the study, which contributed to the

different sample size in each group. Statistics were all done using the SPSS software

package and the significance level was set at P<0.05.

RESULTS

Means and standard deviations for the four groups of athletes are presented in

Table 1. The global test for equality of the mean vectors for the four groups was

significant (Hotelling’s T, P<0.01), which showed that the groups were different. The

Box’s M test was not significant (approximate value of F=1.11, P>0.05), which indicated

homogeneous variance-covariance matrices for each group (Norusis, 1993).

Multiple discriminant analysis revealed three significant functions (P<0.05), DF1,

DF2 and DF3 (Table 2). Based on the Wilk’s Lambda values, DF1 accounted for 78.9%

of the variance while DF2 and DF3 explained 12.8 and 8.3% respectively of the

remaining variance between groups. Standardized coefficients represent an index of the

relative importance of each variable. In other words, standardized coefficients quantify

the potential of each variable for discriminating athletes according to their sport.

Unstandardized coefficients allow the derivation of discriminant scores for each

individual. Based on these scores, group membership could be predicted according to

5
the proximity of the respective group centroid values (mean group values). It is then

possible to determine the percent of correct classification. The higher the percent of

correct classification the more efficient is the model.

As illustrated in Figure 1 and indicated in Table 3, the group centroid distances,

DF1 primarily, describes the anthropometric and biomotor differences between figure

skaters and other groups of athletes. Variables primarily responsible for these disparities

were body mass, height, pushups and maximal girth of the biceps. In contrast, DF2

basically expressed the differences between volleyball players and swimmers. The

variables primarily responsible for these disparities were body mass, maximal girth of the

biceps and calf, and height. All biomotor variables for this function played a secondary

role in explaining the variance between the groups.

The third discriminant function (DF3) illustrated more particularly the differences

between the swimmers and the tennis players. Body mass, biceps and calf girths, sum

of skinfold thicknesses and height best discriminated between these two sport

participants. Although its influence was small, the contribution of the MAP as

discriminator factor was not negligible.

Table 4 summarizes the capability of the model to classify correctly the subjects

in their respective sports. After validation via the jackknife technique, 88% of the athletes

were correctly classified. The probability by chance with four groups was 25%.

DISCUSSION

Most of the variability among groups for DF1 (78.9%) was a function of significant

differences in anthropometric and biomotor parameters between figure skaters and other

groups of athletes. Descriptive data of Table 1 show that figure skaters obtained the

lowest values for all anthropometric variables. This is particularly true for body mass and

height. Figure skaters were smaller and lighter compared to other athletes, despite the

6
lack of a significant difference for age (P>0.05). When compared with a general

population that possessed comparable social and ethnic characteristics (Demirjian et al.,

1972; Léger and Lambert, 1985), it was observed that figure skaters are located at the

20th percentile for height, while athletes from the other sports were generally classified

above the 70th percentile. For body mass, figure skaters were located at the 30th

percentile whereas tennis players, swimmers and volleyball players were respectively

classified at the 55th, 70th and 80th percentiles.

Many explanations are possible for the significant differences found for many

anthropometric values between figure skaters and the other groups of athletes. Some

studies have shown that figure skaters demonstrate a delay (6-12 months) of menarche

compared to the general population (Ross et al., 1976; Malina and Bouchard, 1991).

Such a delay is generally expressed by a lack of development of the body dimensions as

noted by the descriptive data of the present study. We do not know, however, the age of

onset of menarche in our population of athletes. Thus, we cannot confirm in our

population of skaters that their small size is attributable to a lag in physical development

due to a delay in menarche or simply that individuals of that size are naturally attracted

to a discipline that is more suitable to their size In fact, causes that explain this delay are

more dependent of genetic influences and selection processes, rather than training

effects (Stager et al., 1984; Malina, 1994b; Geithner et al., 1998).

From a biomotor point of view, figure skaters reached the best scores for burpees

and trunk flexibility. However, the contribution of these two variables in the model was

modest for all three DF. The burpee scores were not significantly different between any

sports while trunk flexibility was only significant between skaters and tennis players.

These results suggest that the contribution of leg endurance as measured by our test

protocol was similar in these four sports, even if the specificity of the movements is quite

different. In figure skating, leg recruitment is more related to continuous effort when

compared to tennis or volleyball (intermittent) while in swimming, the role of legs is more

7
accessory (Maglisho, 1982). These differences were not apparent based on the score

results. Trunk flexibility differences between skaters and tennis players appear to be

essentially related to the time spent to develop this quality. Figure skaters work daily on

their flexibility, while this was not the case for tennis players. Training effects appear

clearly between these two sports. On the other hand, the smallest value obtained by the

figure skaters for pushups was not surprising since arms are used very little in this sport

compared with the other sports in this study. Thus, the training specificity to a particular

sport may play an important role in defining certain intersport biomotor differences

identified in this study.

The second discriminant function (DF2) primarily described the differences

between swimmers and volleyball players. The standardized coefficients suggest that

body mass, biceps and calf girths and height are the most potent discriminators. Body

mass and height of the swimmers in this study were consistent with the results

presented by Lavoie and Montpetit (1986) with Canadian swimmers of the same age

and performance level. On the other hand, height of the volleyball players agreed with

some other values found in the literature with a certain discrepancy for the body mass. In

his report, Malina (1994a) mentioned that height of volleyball players was located at the

70th percentile compared to the 50th for body mass. In our study, volleyball players were

located at the 80th percentiles for both height and body mass. These results are more

closely related to the values reported for the national American female team (Fleck et al.,

1985).

Although volleyball athletes were taller and heavier, swimmers had a larger bicep

girth, which was a powerful discriminant variable. Mean differences in body mass

between swimmers and volleyball players could be explained partly by the fact that the

latter was also the group who had accumulated the largest quantity of body fat (based

on the sum of the five skinfold thicknesses) among all the sports assessed. The

difference in biceps girth in favour of the swimmers was probably in part a consequence

8
of more intensive muscular training (8 vs 2 h /week for swimmers).

All anthropometric variables with the exception of the humerus and femur

breadth played an important role in explaining the variance between tennis players and

swimmers (DF3). Swimmers were taller, heavier, and presented greater index of

muscularity for the extremity (biceps and calf) than tennis players. Similarly, swimmers

had less body fat and reached better scores for all biomotor variables with the exception

of the MAP, which played a minor role as a discriminator in all three discriminant

functions. This last finding supposes that MAP contributes roughly equally among these

four sports, as shown by their means (Table 1). As was the case for swimmers, height

and body mass of tennis players agreed with values of other investigations (Elliott et al.,

1989; 1990) which may be viewed as consistent with highly proficient adolescent

athletes in that sport.

As indicated by the group centroid values (Table 3), figure skaters presented

physical and biomotor particularities that distinguished them from the other groups. Note

that in Fig. 1 the distance of the centroid group value and the area of the zone for

skaters relative to the other group sports occupy almost half of the total area, while the

other three groups share the remainder. This indicates that the other three groups are

closer to one another when compared to the figure skaters.

Contrary to the other sports, results of this study suggest that small size appears

to be advantageous to reach a high level of performance in figure skating. However, in

order to assess the contribution of the different categories of variables (anthropometric

vs biomotor), discriminant analysis was conducted with anthropometric variables alone

and then with biomotor variables alone (data not shown). As revealed by the discriminant

analysis, anthropomotric variables contributed more to the model than biomotor

variables as discriminators among sports. Thus, when discriminant analysis is repeated

without biomotor variables, 78% of the athletes were correctly classified compared to

9
60% when the anthropometric variables were excluded from the analysis.

The overall percent of correct classification (88%) should be interpreted as very

good and highlights the importance of considering both the anthropometric and the

biomotor variables in the model. This is reinforced by the non-parametric value of

Kappa (0.83), which shows a very good agreement between the observed and predicted

classifications (95% CI: 0.73, 0.92). Table 4 indicates that 100% of the figure skaters

were correctly classified. Moreover, among the athletes misclassified from other sports,

none were allocated as figure skaters. This result confirmed the particular

anthropometric and biomotor profile shown by the figure skaters.

The model appears to be sufficiently sensitive to discriminate among sports. The

general nature of the variables retained allowed to observe differences not related to the

mechanical or physiological specificity of a particular sport. For example, to discriminate

between 100 and 400m swimmers a more sophisticated set of variables would have

been necessary. Nonetheless, in the case of sport categories that are less related to one

another (such as in this study, aquatic, racquet, collective and on-ice sports) the

variables retained are sufficiently powerful to discriminate successfully the athletes

according to their discipline. On a practical point of view, the variables included in the

analysis will allow coaches to use the model easily to assess athlete profiles. An

individual’s high discriminant score, however, in a particular sport does not permit any

conclusion about present or future competing success. Thus, our model will indicate that

a high discriminant score is as an indication that the athlete may possesse a compatible

profile when compared to proficient adolescent athletes within the same sport.

As shown by Pollock et al. (1980), with runners, modifications in the training

process inducing anthropometric or biomotor changes could be quantified by mean of

the discriminant scores and then could indicate if these alterations improved the

probability to classify each individual correctly. Thus, the discriminant analysis could, to

10
some extent, orient the training programmes by identifing the variables affecting the

anthropometric and biomotor profile established by the model.

To our knowledge, there appear to be few, if any, attempts to identify and quantify

variables distinguishing Canadian elite adolescent female athletes from various sports.

Based on the relatively high level of performance reached by the athletes of this study,

the data presented could serve as a guideline for coaches as well as for researchers

looking for a reference model for the four sports investigated.

CONCLUSIONS

Globally, the most potent discriminators among the groups of adolescent female

athletes were body mass, height, biceps and calf girths and muscular endurance

(pushups). These results suggest that each sport is characterized by athletes with

particular anthropometric and biomotor profiles, confirming previous observations

reported in other papers (Pollock et al., 1980;Housh et al., 1984a; 1984c; Watson, 1988).

Despite the relative young age of athletes in the sample (14.3 ± 1.3 years), their profiles

were already well defined. Selection criteria and intensive training effects were possibly

responsible for most of the differences among sports.

11
Acknowledgement. This work was supported in part by the “Fondation des maladies du

coeur du Québec” ASC is a “Fondation de la Recherche en Santé du Québec“ Scholar.

12
 REFERENCES

Callaway, C.W., Chumlea, W.C., Bouchard, C., Himes, J.H., Lohman, T.G., Martin, A.D.,

Mitchell, C.D., Mueller, W.H., Roche A.F. and Seefeldt, V.D.(1988). Circumferences. In

Anthropometric Standardization Reference Manual. T.G. Lohman, A.F. Roche and R.

Martorell (eds), pp. 39-54. Champaign: Human Kinetics Publishers.

Carter, J.E. (1970). The somatotypes of athletes. Human Biology, 42, 553-569.

Carter, J.E. and Ackland, T.R. (1994). Kinanthropometry in Aquatic Sports: a Study of

World Class Athletes. Champaign: Human Kinetics Publisher.

Demirjian, A., Jenicek, M. and Dubuc, M.B. (1972). Les normes staturo-pondérales de

l’enfant urbain canadien français d’âge scolaire. Canadian Journal of Public Health, 63,

14-30.

Elliott, B.C., Ackland, T.R., Blanksby, B.A., Hood, K.P. and Bloomfield, J. (1989).

Profiling junior tennis players part 1: morphological, physiological and psychological

normative data. The Australian Journal of Science and Medicine in Sport, 22, 14-21.

Elliott, B.C., Ackland, T.R., Blanksby, B.A. and Bloomfield, J. (1990). A prospective study

of physiological and kinanthropometric indicators of junior tennis performance. The

Australian Journal of Science and Medicine in Sport, 22, 87-92.

Fleck, S.J., Case, S., Puhl and, J. and Van Handle, P. (1985). Physical and physiological

characteristics of elite women volleyball players. Canadian Journal of Applied Sport

Sciences, 10, 122-126.

Fleishman, E.A. (1965). The Structure and Measurement of Physical Fitness. Englewood

Cliffs, N.J.: Prentice-Hall.

Fogelholm, M., Rankinen, T., Isokaanta, M., Kujala, U. and Uusitupa, M. (2000). Growth,

dietary intake,and trace element status in pubescent athletes and schoolchildren.

Medicine and Science in Sports and Exercise, 32, 738-746.

Geithner, C.A., Woynarowska, B. and Malina R.M. (1998). The adolescent spurt and
sexual maturation in girls active and not active in sport. Annal of Human Biology, 25,

415-423.

Gualdi-Russo, E.G., Gruppioni, G., Gueresi, P., Belcastro, M.G. and Marchesini, V.

(1992). Skinfolds and body composition of sports participants. Journal of Sports

Medicine and Physical Fitness, 32, 303-313.

Housh, T.J., Thorland, W.G., Johnson, G.O. and Tharp, G.D. (1984a). Body build and

compostion variables as discriminators of sports participation of elite adolescent male

athletes. Journal of Sports Medicine and Physical Fitness, 24, 169-174.

Housh, T.J., Thorland, W.G., Johnson, G.O., Tharp, G.D., Cisar, C.J., Refsell, M.J. and

Asorge, C.J. (1984b). Body composition variables as discriminators of sports

participation of elite adolescent female athletes. Research Quaterly for Exercise and

Sport, 55, 302-304.

Housh, T.J., Thorland, W.G., Johnson, G.O., Tharp, G.D. and Cisar, C.J. (1984c).

Anthropometric and body build variables as discriminators of event participation in elite

adolescent male track and field athletes. Journal of Sports Sciences, 2, 3-11.

Lavoie, J.-M. and Montpetit, R.R. (1986). Applied physiology of swimming. Sports

Medicine, 3, 165-189.

Léger, L.A. and Lambert, J. (1985). Poids et taille des québécois de 6 à 17 ans en 1981-

Variations régionales, sexuelles et séculaires. Canadian Journal of Public Health, 76,

388-397.

Leone M. and Léger, L.A. (1985). Épreuves d’endurance musculaire avec rythme

imposé. Canadian Association for Physical Health, Exercise and Recreation, 55, 27-33.

Loko, J., Aule, R., Sikkut, T., Ereline, J. and Viru, A. (2000). Motor performance status in

10 to 17-year-old Estonian girls. Scandinavia Journal of Medicine and Science in Sports,

10, 109-113.

MacDougall, J.D., Wenger, H.A. and Green, H.J. (1988). Évaluation Physiologique de
l’Athlète de Haut Niveau. Montréal: Décarie-Vigot.

Maglisho, E.W. (1982). Swimming Faster. California: Mayfield Publishing Company.

Malina R.M. (1994a). Attained size and growth rate of female volleyball players between

9 and 13 years of age. Pediatric Exercise Science, 6, 257-266.

Malina, R.M. (1994b). Physical activity and training: effects on stature and the

adolescent growth spurt. Medicine and Science in Sports and Exercises, 26, 759-766.

Malina, R.M. and Bouchard, C. (1991). Characteristics of young athletes. In: Growth,

Maturation, and Physical Activity, pp. 443-463, Champaign, Il. Human Kinetics.

Norusis, M.J. (1993). Discriminant analysis. In: SPSS for Windows Release 6.0:

Professional Statistics, pp.1-45. Chicago, Il: SPSS Inc.

Orvana, E. (1987). Physical structure of winter sports athletes. Journal of Sports

Sciences, 5, 197-248.

Pollock, M.L., Jackson, A.S. and Pate, R.R. (1980). Discriminant analysis of

physiological differences between good and elite distance runners. Research Quarterly

for Exercise and Sport, 51, 521-532.

Ross, W.D., Brown, S.R., Faulker, R.A. and Savage, M.V. (1976). Age of menarche of

elite canadian skaters and skiers. Canadian Journal of Applied Sport Sciences, 1, 191-

193.

Smith, H.K. and Thomas, S.G. (1991). Physiological characteristics of elites female

basketball players. Canadian Journal of Sport Sciences, 16, 289-295.

Sovak, D. and Hawes, M.R. (1987). Anthropological status of international calibre speed

skaters. Journal of Sports Sciences, 5, 287-304.

Stager, J.M., Robertshaw, D. and Miescher, E. (1984). Delayed of menarche in

swimmers in relation to age at onset of training and athletic performance. Medicine and

Science in Sports and Exercise 16:550-555.

Thomas, S.G., M.H. Cox, Y.M. LeGal, T.J. Verde and H.K. Smith (1989). Physiological
profiles of canadians national judo team. Canadian Journal of Sport Sciences, 14, 142-

147.

Watson, A.W.S. (1988) Discriminant analysis of the physiques of schoolboy rugby

players and non-team members. Journal of Sports Sciences, 6, 131-140.

Wilmore, J.H., R.A. Frisancho, C.C. Gordon, J.H. Himes, A.D. Martin, R. Martorell and

V.D. Seefeldt (1988). Body breadth equipment and measurement techniques.In:

Anthropometric Standardization Reference Manual, (edited byT.G. Lohman, A.F. Roche

and R. Marotell). Champaign, Il.: Human Kinetics, chap. 3, pp. 27-38.

Figure Legend

Fig. 1. Territorial map of the athletes tested relative to their discipline. The asterisks

indicates the group centroid (Volleyball (Volley), Tennis; Swimming; and Skating). DF1,

discriminant function 1; and DF2, discriminant function 2.

Table 1. Descriptive (mean and standard deviation) characteristics for athletes in each sport.

Tennis Skating Swimming Volleyball

(n=15) (n=46) (n=23) (n=16)

Age (yrs) 13.9±1.3 14.7±1.5 14.3±1.3 13.8±1.3

Body mass (Kg) 50.6±8.3 46.6±8.0 54.3±6.9 57.7±8.3

Height (cm) 160.6±6.0 153.8±6.5 162.0±5.7 163.1±5.4

Elbow (cm) 6.12±0.30 5.87±0.35 6.29±0.26 6.40±0.33

Knee (cm) 8.81±0.43 8.63±0.76 8.77±0.34 9.31±0.50

Biceps (cm) 25.5±2.8 24.4±2.3 27.8±1.8 26.6±2.2

Calf (cm) 34.0±2.8 33.0±2.7 34.4±1.6 34.4±2.2

Skinfolds (mm) 57.4±17.8 47.7±12.3 56.0±15.0 63.1±15.5

Push-up (number) 57.8±14.4 36.7±13.5 62.1±16.0 50.2±13.5

Burpee (number) 46.1±23.8 64.6±33.2 52.5±32.7 56.0±28.4

Flexibility (cm) 37.3±5.0 42.6±5.1 41.0±6.0 39.1±6.9

VO2max (ml.kg. -1.min-1) 49.5±4.4 48.3±4.0 47.6±3.1 48.9±3.6

Table 2. Statistical significance and standardized and unstandardized coefficients for
each discriminant function.

Standardized coefficients Unstandardized coefficients

Variables

DF1 DF2 DF3 DF11 DF22 DF33

Age -0.627 -0.330 0.347 -0.442 -0.233 0.245

Body mass -1.291 2.309 -2.001 -0.164 0.294 -0.255

Height 1.554 -0.505 0.867 0.254 -0.083 0.142

Elbow 0.154 0.211 0.042 0.476 0.653 0.129

Knee -0.599 0.415 -0.256 -0.995 0.689 -0.425

Biceps 0.729 -1.532 -1.066 0.326 -0.686 -0.477

Calf 0.333 -0.495 1.785 0.136 -0.203 0.730

Skinfolds 0.440 0.043 0.984 0.031 0.001 0.068

Push-up 0.797 -0.015 0.132 -0.015 0.001 -0.001

Burpee -0.462 0.300 -0.225 -0.015 0.001 -0.001

Flexibility -0.326 -0.304 -0.300 -0.058 -0.054 -0.053

VO2max 0.012 0.057 0.439 0.000 0.015 0.116

(Constant) -33.945 16.713 -29.513

1
DF1: eigenvalues of BW-1 (V = 221.3 d.f. = 36 P<0.05)
2
DF1: First Residual (V1 = 76.3 d.f. = 22 P<0.05)
3
DF1: Second Residual (V2 = 31.4 d.f. = 10 P<0.05)

Hence, all three DF are significant, since the eigenvalues of BW -1 (ratio of between and within

group variation), the first residual (V1) and the second residual (V2) are all significant at 5%

level.
Table 3. Group centroids for each discriminant function.

Group Centroids

Sports DF1 DF2 DF3

Tennis 1.8970 0.0290 1.3657

Skating -2.0539 -0.1784 0.0129

Swimming 2.2327 -0.8197 -0.6054

Volleyball 0.9169 1.6641 -0.4470

Table 4. Classification for all significant discriminant functions after validation.

Predicted group membership – n – (percent)

Groups n 1 2 3 4

1. Tennis 15 11(73.3) 0(0.0) 3(20.0) 1(6.7)

2. Skating 46 0(0.0) 46(100) 0(0.0) 0(0.0)

3. Swimming 23 1(13.0) 0(0.0) 18(78.3) 2(8.7)

4. Volleyball 16 1(6.3) 0(0.0) 2(12.6) 13(81.3)

Percent of correctly classified cases: 88.0% (P<0.05).