lable at ScienceDirect

Applied Ergonomics 42 (2010) 156e161

Contents lists avai

Applied Ergonomics

journal homepage: www.elsevier .com/locate/apergo

Evaluation of the multivariate accommodation performance of the grid method

Kihyo Jung a, Ochae Kwon b, Heecheon You c,*

a The Harold and Inge Marcus Department of Industrial and Manufacturing Engineering, The Pennsylvania State University, University Park, PA 16802, USAb Samsung Electronics Co., Ltd. 10F, Samsung Electronics Building, Seocho 2-dong, Seocho-gu, Seoul 137-857, South KoreacDepartment of Industrial and Management Engineering, Pohang University of Science and Technology, Pohang, Kyungbuk 790-784, South Korea

a r t i c l e i n f o

Article history:Received 9 October 2009Accepted 18 June 2010

Keywords:Multivariate accommodation performanceGrid methodRepresentative human model generationAnthropometry

* Corresponding author. Tel.: þ82 54 279 2210; faxE-mail address: hcyou@postech.ac.kr (H. You).

0003-6870/$ e see front matter � 2010 Elsevier Ltd.doi:10.1016/j.apergo.2010.06.004

a b s t r a c t

The present study examined the multivariate accommodation performance (MAP) of the grid method,a distributed representative human models (RHM) generation method, in the context of men’s pantssizing system design. Using the 1988 US Army male anthropometric data and �2.5 cm of fitting toler-ance, the grid method selected two key dimensions (waist girth and crotch height) out of 12 anthro-pometric dimensions and identified 25 RHMs to accommodate 95% of the population. The average MAPof the RHMs decreased dramatically as the number of anthropometric dimensions considered increased(99% for single dimension and 14% for 12 dimensions). A standardized regression model was establishedwhich explains the effects of two factors (sum of anthropometric dimension ranges; adjusted R2 betweenkey dimensions and other anthropometric dimensions) on the MAP of RHMs. This regression model canbe used to prioritize anthropometric dimensions for efficient MAP improvement of men’s pants design.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

In ergonomic design and evaluation, a small group of humanmodels which represents the anthropometric variability of thetarget population is commonly used. Representative humanmodels (RHMs) are determined by considering the anthropometriccharacteristics of the population and a designated accommodationpercentage (e.g., 90%) (HFES 300, 2004; Jung et al., 2009). Forexample, You et al. (1997) used three RHMs (5th, 50th, and 95thpercentiles) which accommodate 90% of the target population toevaluate an ergonomic design of a bus operator’s workstation. Useof a small group of RHMs provides an efficient way to developa product designwhich is fitting for the population. However, somestudies use a rather large group of human models (e.g., 6e45manikins) to better reflect the diversity of anthropometric sizes forergonomic design and evaluation (Bittner, 2000; Case et al., 2001,2009; Jung et al., 2009).

Depending on the characteristics of a product being designed,RHMs are determined on a boundary (Fig. 1a) or a set of grids(Fig. 1b) formed on selected anthropometric dimensions toaccommodate a designated percentage of the population. Indesigning a one-size product (one-size to accommodate peoplewithin a designated percentage of the population) such as a bus

: þ82 54 279 2870.

All rights reserved.

operator’s workstation and a helicopter cockpit, RHMs are selectedon an accommodation boundary. For example, Bittner (2000)generated 17 RHMs at the centroid and the boundary of a hyper-ellipsoidwhich encompasses 90% of the population for aworkspacedesign. On the other hand, in designing a multiple-size product (nsizes to fit n groups of people within a designated percentage of thepopulation) such as gloves and shoes, RHMs are chosen over a set ofgrids formed in the distribution of anthropometric dimensions. Forexample, Kwon et al. (2009) and Robinette and Annis (1986) con-structed grids which accommodated a designated percentage of thepopulation and then defined representative cases at the centroidsof the grids.

Of the various RHM generation methods, the grid methoddetermines RHMs by following a three-step procedure: selectionof key dimensions, formation of representative grids, and gener-ation of RHMs (Fig. 2). First, a small, manageable number ofanthropometric dimensions (e.g., one to five dimensions) areselected as key dimensions by analyzing the statistical relation-ships between anthropometric dimensions under consideration(Gordon and Friedl, 1994; Hidson, 1991; Rosenblad-Wallin, 1987;Zheng et al., 2007). Second, representative grids with a fittingtolerance are formed over the distribution of the key dimensionsto accommodate a designated percentage of the population. Thevalue of fitting tolerance is specified by the product designer byconsidering various practical aspects including productioneconomy and material properties (Kwon et al., 2009). Lastly, thevalues of the centroid of each representative grid are used for the

Fig. 1. Determination of representative cases (small dots: population cases; large dots: representative cases).

K. Jung et al. / Applied Ergonomics 42 (2010) 156e161 157

key dimensions of an RHM and then regression equations havingthe key dimensions as regressors are applied to estimate theother anthropometric dimensions of the RHM (Robinette andAnnis, 1986).

Research which evaluates the accommodation performance ofthe grid method in a comprehensive manner is lacking. Moststudies such as Chung et al. (2007) and McCulloch et al. (1998)using the grid method have considered the multivariate accom-modation performance (MAP) of the grid method only for keydimensions, but not for non-key anthropometric dimensions. Sincesome non-key anthropometric dimensions have low correlationswith key dimensions but still affect the fit of product to the pop-ulation, the MAP of the grid method needs to be analyzed for allanthropometric dimensions considered in the design.

The present study examined the MAP of the grid method ina comprehensive manner and the effects of two factors (sum ofanthropometric dimension ranges; adjusted R2 between keydimensions and other anthropometric dimensions) on MAP. Themeasure MAP was quantified by calculating the proportion of thepopulation belonging to grids formed to accommodate a desig-nated percentage of the population. The two MAP factors wereidentified from our empirical observation as possible attributesaffecting the MAP of the grid method. Lastly, standardized multipleregression analysis was conducted to examine the relative effects ofthe two factors on MAP.

Fig. 2. Representative human model (RHM) generation process of the

2. Methods

2.1. Selection of anthropometric dimensions

Referring to Chung et al. (2007) and Hsu and Wang (2005), thepresent study selected 12 anthropometric dimensions (Table 1) forthe design of a men’s pants sizing system. The 1988 US Armyanthropometric data of 1774 men (Gordon et al., 1988) was used togenerate RHMs. The anthropometric data was randomly dividedinto (1) a learning data set of 1000 cases for generation of RHMsand (2) a testing data set of 774 cases for cross-evaluation of theaccommodation performance of the generated RHMs.

2.2. Determination of key dimensions

To determine the number of key dimensions for a men’s pantssizing system, maximum average adjusted R2 was analyzed fordifferent numbers of key dimension candidates as displayed inFig. 3. The 12 anthropometric dimensions were used to form sets ofkey dimension candidates consisting of 1e11 anthropometricdimensions. Then, multiple regression analysis was conductedusing each set of key dimension candidates as regressors and non-key anthropometric dimensions as dependent variables. Finally, foreach number of key dimension candidates, the maximum ofaverage adjusted R2 values were identified. Table 2 illustrates the

grid method (AD: anthropometric dimension; K: key dimension).

Table 1Anthropometric dimensions selected for design of men’s pants sizing system.

Dimensionaltype

Code Selecteddimensions

Hsu andWang (2005)

Chunget al. (2007)

Descriptive statisticsb

Mean SD Range

Lengthand height

AD1 Waist height B 112.7 5.2 43.1AD2 Waist-to-ankle lengtha B 106.0 5.0 40.3AD3 Waist-to-knee lengtha B 56.8 2.7 21.0AD4 Outside leg length B 108.2 5.1 41.0AD5 Crotch height B B 83.7 4.6 39.2AD6 Crotch length B B 76.7 5.6 38.1

Girth AD7 Waist girth B B 84.0 7.4 47.5AD8 Hip girth B 98.4 6.2 43.4AD9 Thigh girth B B 59.7 4.9 32.9AD10 Knee girth B 38.6 2.2 14.7AD11 Calf girth B 37.8 2.5 16.6AD12 Ankle girth B 22.2 1.3 9.0

a Waist-to-ankle length and waist-to-knee length were estimated by subtracting ankle height and knee height from waist height, respectively, because of their unavail-ability in the 1988 US Army anthropometric data.

b Source: The 1988 US Army anthropometric data of males (n ¼ 1774) (Gordon et al., 1988).

Table 2Analysis of average adjusted R2 for single key dimension candidates (illustrated).

Key dimension Anthropometric Adjusted R2 Average2

K. Jung et al. / Applied Ergonomics 42 (2010) 156e161158

identification process of the maximum of average adjusted R2

values for single key dimension candidate cases: for each individualkey dimension candidate from AD1 to AD12, adjusted R2 values ofregression equations were obtained for the other anthropometricdimensions and their average adjusted R2 value was calculated, andthen themaximum of the adjusted R2 values was found 0.43 at AD1.Fig. 3 shows that the increase rate of maximum average adjusted R2

becomes significantly small (<0.1) after two; thus, this studydetermined that two anthropometric dimensions would beselected for key dimensions of a men’s pants sizing system.

Based on the maximum average adjusted R2 analysis results anda survey on key dimensions used in the garment industry, crotchheight (AD5) and waist girth (AD7) were chosen as key dimensionsof a men’s pants sizing system design. The key dimension surveyconducted on the internet in the present study identified that AD5and AD7 are widely used for men’s pants sizing in the US garmentindustry (Banana Republic, 2009; Polo, 2009; Perry Ellis, 2009).

2.3. Calculation of multivariate accommodation performance

The MAP of a particular set of anthropometric dimensions wasquantified by calculating the proportion of anthropometric casesbelonging to grids that were formed along generated RHMs. Afitting tolerance of�2.5 cmwas used as the length of a grid for eachanthropometric dimension by referring to the garment industrypractices for the key dimensions of men’s pants (Banana Republic,

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1 2 3 4 5 6 7 8 9 10 11

2R detsujda fo egarev

A

Number of key dimensions

detsuj

da mu

mixaM

R2

snoisnemi

d rehto htiw

Fig. 3. The trend of maximum adjusted R2 by the number of key dimensions.

2009; BlueFly, 2009; Polo, 2009; Perry Ellis, 2009). The MAP anal-ysis was conducted for various combinations of anthropometricdimensions from 1 to 12 using a program coded by Matlab 7.0(MathWorks, Inc., Natick, MA, USA). For example, when the numberof anthropometric dimensions was 2, MAPs were calculated for 66combinations (12C2) of anthropometric dimensions.

3. Results

Using the grid method, 25 RHMs (Table 3) were generated forthe design of a men’s pants sizing system to accommodate 95% ofthe population. Square grids with a fitting tolerance of �2.5 cmwere formed which accommodate 95% of the population anthro-pometric cases (Fig. 4) on the distribution of AD5 and AD7. Thecentroids of the grids were used for the key dimension values of theRHMs, and then the regression equations in Table 4, having the keydimensions as regressors, were applied to determine the values ofthe other anthropometric dimension values for the RHMs.

The MAP analysis (described in the Methods section) showedthat the average MAP of the generated RHMs decreased greatly asthe number of anthropometric dimensions increased as displayed

candidate dimensions adjusted R

AD1 AD2AD3AD4AD5AD6AD7AD8AD9AD10AD11AD12

0.990.890.880.830.200.100.180.120.230.120.15

0.43

AD2 AD1AD3«

AD12

0.990.89«

0.13

0.42

« « « «

AD12 AD1AD2«

AD11

0.150.13«

0.64

0.28

Table 3Representative human models generated by the grid method to accommodate 95% of the 1988 US Army male population.a

No Key dimensions Non-key anthropometric dimensionsb

Crotchheight(AD5)

Waistgirth(AD7)

Waistheight(AD1)

Waist-to-anklelength (AD2)

Waist-to-kneelength (AD3)

Outside leglength (AD4)

Crotchlength(AD6)

Hipgirth(AD8)

Thighgirth(AD9)

Kneegirth(AD10)

Calfgirth(AD11)

Anklegirth(AD12)

1 73.8 (2) 74.1 (9) 101.0 (1) 94.8 (1) 51.3 (2) 96.8 (1) 69.9 (11) 89.5 (8) 53.5 (11) 35.5 (8) 34.9 (13) 20.7 (14)2 73.8 (2) 79.1 (25) 101.9 (2) 95.6 (2) 51.8 (3) 97.6 (2) 72.9 (25) 93.0 (19) 56.1 (24) 36.5 (17) 36.0 (24) 21.2 (22)3 73.8 (2) 84.1 (50) 102.8 (3) 96.4 (3) 52.3 (5) 98.4 (3) 75.8 (44) 96.4 (38) 58.7 (42) 37.5 (31) 37.1 (39) 21.6 (34)4 78.8 (15) 69.1 (2) 105.2 (7) 98.9 (8) 53.2 (9) 100.9 (8) 67.4 (5) 87.0 (3) 51.4 (5) 35.0 (5) 34.2 (8) 20.6 (11)5 78.8 (15) 74.1 (9) 106.0 (10) 99.7 (10) 53.7 (12) 101.7 (10) 70.3 (13) 90.5 (10) 54.0 (13) 36.0 (12) 35.3 (16) 21.0 (19)

6 78.8 (15) 79.1 (25) 106.9 (13) 100.5 (14) 54.1 (16) 102.6 (13) 73.3 (27) 94.0 (24) 56.6 (27) 37.1 (24) 36.4 (29) 21.5 (29)7 78.8 (15) 84.1 (50) 107.8 (17) 101.3 (17) 54.6 (21) 103.4 (17) 76.3 (47) 97.4 (44) 59.2 (46) 38.1 (40) 37.5 (45) 21.9 (42)8 78.8 (15) 89.1 (75) 108.7 (22) 102.1 (22) 55.0 (26) 104.2 (22) 79.2 (68) 100.9 (66) 61.8 (67) 39.1 (59) 38.6 (62) 22.3 (55)9 78.8 (15) 94.1 (91) 109.6 (27) 102.9 (27) 55.5 (31) 105.1 (27) 82.2 (84) 104.4 (83) 64.4 (83) 40.2 (75) 39.7 (77) 22.8 (68)10 83.8 (51) 69.1 (2) 110.2 (31) 103.7 (32) 55.5 (32) 105.8 (32) 67.8 (6) 88.0 (5) 51.9 (6) 35.6 (8) 34.6 (10) 20.9 (16)

11 83.8 (51) 74.1 (9) 111.1 (38) 104.5 (38) 56.0 (38) 106.7 (38) 70.8 (14) 91.5 (13) 54.5 (15) 36.6 (18) 35.6 (20) 21.3 (25)12 83.8 (51) 79.1 (25) 112.0 (44) 105.3 (45) 56.4 (44) 107.5 (45) 73.7 (30) 95.0 (29) 57.1 (30) 37.6 (33) 36.7 (34) 21.7 (37)13 83.8 (51) 84.1 (50) 112.9 (51) 106.1 (51) 56.9 (51) 108.4 (51) 76.7 (50) 98.4 (50) 59.7 (50) 38.7 (50) 37.8 (50) 22.2 (50)14 83.8 (51) 89.1 (75) 113.7 (58) 106.9 (57) 57.4 (58) 109.2 (58) 79.7 (71) 101.9 (72) 62.3 (70) 39.7 (68) 38.9 (67) 22.6 (64)15 83.8 (51) 94.1 (91) 114.6 (64) 107.8 (64) 57.8 (64) 110.0 (64) 82.6 (86) 105.4 (87) 64.9 (86) 40.7 (83) 40.0 (81) 23.1 (75)

16 83.8 (51) 99.1 (98) 115.5 (70) 108.6 (70) 58.3 (70) 110.9 (70) 85.6 (95) 108.8 (95) 67.5 (94) 41.8 (92) 41.1 (90) 23.5 (85)17 88.8 (87) 74.1 (9) 116.1 (74) 109.4 (75) 58.3 (70) 111.6 (75) 71.2 (16) 92.5 (17) 55.0 (17) 37.2 (25) 36.0 (24) 21.6 (33)18 88.8 (87) 79.1 (25) 117.0 (80) 110.2 (80) 58.8 (76) 112.5 (80) 74.2 (33) 96.0 (35) 57.6 (34) 38.2 (42) 37.1 (39) 22.0 (46)19 88.8 (87) 84.1 (50) 117.9 (84) 111.0 (84) 59.2 (81) 113.3 (84) 77.2 (53) 99.4 (57) 60.2 (54) 39.2 (61) 38.2 (56) 22.5 (59)20 88.8 (87) 89.1 (75) 118.8 (88) 111.8 (88) 59.7 (85) 114.2 (88) 80.1 (73) 102.9 (77) 62.8 (74) 40.3 (77) 39.3 (72) 22.9 (71)

21 88.8 (87) 94.1 (91) 119.7 (91) 112.6 (91) 60.1 (89) 115.0 (91) 83.1 (88) 106.4 (90) 65.4 (88) 41.3 (88) 40.4 (84) 23.3 (82)22 88.8 (87) 99.1 (98) 120.5 (93) 113.4 (93) 60.6 (92) 115.8 (93) 86.1 (95) 109.8 (97) 68.0 (95) 42.3 (95) 41.5 (92) 23.8 (89)23 93.8 (99) 79.1 (25) 122.0 (96) 115.0 (96) 61.1 (94) 117.4 (96) 74.6 (36) 97.0 (41) 58.1 (38) 38.8 (53) 37.5 (45) 22.3 (54)24 93.8 (99) 84.1 (50) 122.9 (98) 115.8 (98) 61.5 (96) 118.3 (98) 77.6 (57) 100.5 (63) 60.7 (58) 39.8 (70) 38.5 (61) 22.7 (67)25 93.8 (99) 89.1 (75) 123.8 (98) 116.6 (98) 62.0 (97) 119.1 (98) 80.6 (76) 103.9 (81) 63.3 (77) 40.8 (84) 39.6 (76) 23.2 (78)

a The values are in cm and corresponding percentile values are presented in parentheses.b Non-key dimensions were estimated by regression equations in Table 4.

K. Jung et al. / Applied Ergonomics 42 (2010) 156e161 159

in Fig. 5. The univariate accommodation percentage of each indi-vidual anthropometric dimension was higher (96.6%e99.9%) thanthe target accommodation percentage (95%); however, the averageMAP decreased dramatically as the number of anthropometricdimensions increased and finally dropped to 14% when all 12dimensions were considered.

During the process of MAP analysis, two factors (sum ofanthropometric dimension ranges, SR; average adjusted R2

Fig. 4. Formation of representative grids (fitting tolerance ¼ �2.5 cm) accommodating95% of the 1988 US Army male population (small dots: population cases; large dots:representative cases) (illustrated).

between key dimensions and other anthropometric dimensions,AR) were presumed as those affecting the MAP of the grid method.It was observed that SR tended to negatively relate to MAP. Forexample, the MAP of waist-to-knee length (AD3) (range ¼ 21 cm),waist girth (AD7) (range ¼ 47.5 cm), and ankle girth (AD12)(range¼ 9 cm), of which SR¼ 77.5 cm, was 94%, while that of crotchlength (AD6) (range ¼ 38.1 cm), hip girth (AD8) (range ¼ 43.4 cm),and thigh girth (AD9) (range ¼ 32.9 cm), of which SR ¼ 104.4 cm,was 70% (note that the range data are from Table 1). In contrast, itwas observed that AR tended to positively relate to MAP. Forexample, the MAP of waist height (AD1) (adjusted R2 ¼ 0.89), AD3(adjusted R2 ¼ 0.70), and outside leg length (AD4) (adjustedR2 ¼ 0.90), of which AR ¼ 0.84, was 95%, while that of waist-anklelength (AD2) (adjusted R2 ¼ 0.88), AD6 (adjusted R2 ¼ 0.64), andAD9 (adjusted R2 ¼ 0.62), of which AR ¼ 0.71, was 64%.

The statistical significance and relative influence of the twoMAPfactors were further examined by multiple regression analysis. Astandardized regression equation (F(2, 217) ¼ 95.4, p < 0.001;adjusted R2 ¼ 0.46) was obtained by the stepwise approach(probabilities to enter and to remove ¼ 0.05 and 0.1):

MAPð%Þ ¼ ð1:03� 0:616� SR þ 0:284� ARÞ � 100 (1)

The regression coefficients indicating the directionalities of thefactor effects agreed with the aforementioned observations: SRnegatively relates to MAP, while AR positively relates to MAP. Also,the standardized regression equation showed that SR is moreinfluential to MAP.

4. Discussion

The present study examined the MAP of the grid method for allthe combinations of both key and non-key anthropometric

Table 4Regression equations using crotch height (AD5) and waist girth (AD7) as regressors.

Anthropometric dimensions Code Regression equation Adjusted R2

Waist height AD1 AD1 ¼ 13.49 þ 1.01 � AD5 þ 0.18 � AD7 0.89Waist-to-ankle length AD2 AD2 ¼ 11.24 þ 0.97 � AD5 þ 0.16 � AD7 0.88Waist-to-knee length AD3 AD3 ¼ 10.33 þ 0.46 � AD5 þ 0.09 � AD7 0.70Outside leg length AD4 AD4 ¼ 11.07 þ 0.99 � AD5 þ 0.17 � AD7 0.90Crotch length AD6 AD6 ¼ 19.53 þ 0.09 � AD5 þ 0.59 � AD7 0.64Hip girth AD8 AD8 ¼ 23.25 þ 0.20 � AD5 þ 0.69 � AD7 0.72Thigh girth AD9 AD9 ¼ 7.76 þ 0.10 � AD5 þ 0.52 � AD7 0.62Knee girth AD10 AD10 ¼ 11.77 þ 0.11 � AD5 þ 0.21 � AD7 0.55Calf girth AD11 AD11 ¼ 13.47 þ 0.07 � AD5 þ 0.22 � AD7 0.43Ankle girth AD12 AD12 ¼ 10.04 þ 0.06 � AD5 þ 0.09 � AD7 0.30

K. Jung et al. / Applied Ergonomics 42 (2010) 156e161160

dimensions. Previous research has evaluated the MAP of the gridmethod only for key anthropometric dimensions (Chung et al.,2007; McCulloch et al., 1998), but not for non-key anthropometricdimensions. Although key dimensionsmainly determine the fitnessof a product and an appropriate size of a product for the user (Kwonet al., 2009), non-key anthropometric dimensions still affect itsfitness; thus, an understanding of the accommodation performanceof the grid method is necessary for non-key anthropometricdimensions.

The MAP analysis results in the present study showed that thegrid method can over- and under-accommodate a designatedaccommodation percentage of the population depending onanthropometric dimensions considered. In addition, relatively largedecrease rates (>10%) or variabilities (>20%) in accommodationperformance were observed in the mid range of the number ofanthropometric dimensions (3e8) in which the MAPs for all thecombinations of the 12 anthropometric dimensions were analyzedin the study.

The MAP quantification method used in the present study isapplicable to accommodation performance evaluation of any RHMgeneration methods. Previous research has developed variousstatistical and optimization methods to generate RHMs over thetarget population distribution. For example, Eynard et al. (2000)and Laing et al. (1999) used cluster analysis to classify the targetpopulation into representative figure types and then generateRHMs for each figure type; McCulloch et al. (1998) applied anoptimization algorithm under given constraints (e.g., number ofRHMs and fitting tolerance) to generate RHMs. However, compar-ison between RHM generation methods has not been made interms ofMAP. An in-depthMAP analysis is required in the future forexisting RHM generation methods.

99 93

82

69

57

46

37 30

24 20

16 140

20

40

60

80

100

1 2 3 4 5 6 7 8 9 10 11 12

)%( noitado

mmocc

A

Number of anthropometric dimensions

Fig. 5. Average and range (min and max) of accommodation percentage by thenumber of anthropometric dimensions.

The value of fitting tolerance in MAP analysis is often deter-mined by considering product fitness and production economy(Kwon et al., 2009; McCulloch et al., 1998). A small value of fittingtolerance can increase the level of fit for a product to the users, butrequires a large number of size categories which negatively affectsproduction economy; the opposite becomes true for a large value offitting tolerance.

Different values of fitting tolerance can be used in MAP analysis.The present study applied �2.5 cm as a fitting tolerance uniformlyto all the anthropometric dimensions for illustration purposes.However, a value of fitting tolerance for a key or non-key anthro-pometric dimension would differ depending on specific designconditions such as material properties, production economyconsiderations, wearing purposes, and design concepts of clothing.Thus, various values of fitting tolerance need to be applied toanthropometric dimensions when applying the proposed MAPanalysis procedure by considering the design context.

A standardized multiple regression model was established toexamine the effects of SR and AR on MAP. The MAP regressionmodel showed that the less the SR, the higher the MAP and that theopposite is held for AR. The MAP regression model can be used toanalyze the prioritization of design improvements for betteraccommodation performance of a product design. For example,material changes or adjustment mechanisms can be examinedwitha higher priority for design parameters that are related to anthro-pometric dimensions having a larger SR and/or a smaller AR. Toobtain a more comprehensive, optimal solution for designimprovement in industry, cost factors would be taken into accountin addition to MAP sensitivity results.

Acknowledgments

This research was supported by Basic Science Research Programthrough the National Research Foundation of Korea (NRF) fundedby the Ministry of Education, Science, and Technology (2010-0012291).

References

Banana Republic, 2009. Men’s pant sizing system. Retrieved 10.07.09. from. http://www.gap.com/browse/category.do?cid¼11889.

Bittner, A.C., 2000. A-CADRE: advanced family of manikins for workstation design.In: Proceedings of the IEA 2000/HFES 2000 Congress, San Diego, CA,pp. 774e777.

BlueFly, 2009. Men’s pant sizing system. Retrieved 10.07.09. from. http://www.bluefly.com/Theory-wheat-khaki-wool-MarloWythe-trousers/cat60015/304852501/detail.fly.

Case, K., Marshall, R., Hogberg, D., Summerskill, S., Gyi, D., Sims, R., 2009. HADRIAN:fitting trials by digital human modelling. Lecture Notes in Computer Science,673e680.

Case, K., Porter, M., Gyi, D., Marshall, R., Oliver, R., 2001. Virtual fitting trials indesign for all. Journal of Materials Processing Technology 117, 255e261.

K. Jung et al. / Applied Ergonomics 42 (2010) 156e161 161

Chung, M., Lin, H., Wang, M., 2007. The development of sizing systems for Taiwa-nese elementary- and high-school students. International Journal of IndustrialErgonomics 37 (8), 707e716.

Eynard, E., Fubini, E., Masali, M., Cerrone, M., Tarzia, A., 2000. Generation of virtualman models representative of different body proportion and application toergonomic design of vehicles. In: Proceedings of the IEA2000/HFES2000Congress, San Diego, CA, pp. 489e492.

Gordon, C.C., Bradtmiller, B., Churchill, T., Clauser, C., McConville, J., Tebbetts, I.,Walker, R., 1988. 1988 Anthropometric Survey of US Army Personnel: Methodsand Summary Statistics (Technical Report NATICK/TR-89/044). US Army NatickResearch Center, Natick, MA.

Gordon, C.C., Friedl, K.E., 1994. Anthropometry in the US armed forces. In:Ulijaszek, S.J. (Ed.), Anthropometry: The Individual and the Population.Cambridge University Press, Cambridge UK.

HFES 300, 2004. Guidelines for Using Anthropometric Data in Product Design.Human Factors and Ergonomics Society, Santa Monica, California.

Hidson, D., 1991. Development of a Standard Anthropometric Dimension Set for Usein Computer-Aided Glove Design (DREO Technical Note 91-22). DefenseResearch Establishment, Ottawa.

Hsu, C., Wang, M., 2005. Using decision tree-based data mining to establish a sizingsystem for the manufacture of garments. International Journal of AdvancedManufacturing Technology 26, 669e674.

Jung, K., Kwon, O., You, H., 2009. Development of a digital human model generationmethod for ergonomic design in virtual environment. International Journal ofIndustrial Ergonomics 39 (5), 744e748.

Kwon, O., Jung, K., You, H., Kim, H., 2009. Determination of key dimensions fora glove sizing system by analyzing the relationships between hand dimensions.Applied Ergonomics 40 (4), 762e766.

Laing, R.M., Holland, E.J., Wilson, C.A., Niven, B.E., 1999. Development of sizingsystems for protective clothing for the adult male. Ergonomics 42 (10),1249e1257.

McCulloch, C.E., Paal, B., Ashdown, S.P., 1998. An optimization approach to apparelsizing. Journal of the Operational Research Society 49 (5), 492e499.

Polo, 2009. Men’s pant sizing system. Retrieved 10.07.09. from. http://www.ralphlauren.com/product/index.jsp?productId¼3678183&cp¼1760781.1760811&ab¼ln_men_cs1_pants&parentPage¼family.

Perry Ellis, 2009. Men’s pant sizing system. Retrieved 10.07.09. from. http://www.perryellis.com/Pants/Micro-Twill-Pant/invt/49mb8308ps&bklist¼icat,4,shop,pemen,pedmbott.

Robinette, K.M., Annis, J.F., 1986. A Nine-Size System for Chemical Defense Gloves.Harry G. Armstrong Aerospace Medical Research Laboratory, Wright-PattersonAir Force Base, OH. Technical Report (AAMRL-TR-86e029) (ADA173 193).

Rosenblad-Wallin, E., 1987. An anthropometric study as the basis for sizinganatomically designed mittens. Applied Ergonomics 18 (4), 329e333.

You, H., Bucciaglia, J., Lowe, B.D., Gilmore, B.J., Freivalds, A., 1997. An ergonomicdesign process for a US transit bus operator workstation. Heavy VehicleSystems, A Series of the International Journal of Vehicle Design 4 (2e4), 91e107.

Zheng, R., Yu, W., Fan, J., 2007. Development of a new Chinese bra sizing systembased on breast anthropometric measurements. International Journal ofIndustrial Ergonomics 37 (8), 697e705.