1 The adaptation of Gilliam Autism Rating Scale-2 within an Omani

2 context: Some initial findings

Aim: The aim of the present study was to report some initial findings concerning the validity and

reliability of the Omani version of the Gilliam Autism Rating Scale-2 (OM-GARS-2).

Sample: The sample of the study included 90 children aged 8-14 years and divided into two

groups: Autistic group (n = 45), enrolled in two public centers of autism care in two governorates

in Oman, and normal group (n =45), enrolled in two public schools in two governorates in

Oman.

Raters sample. A total of 8 teachers (4 males and 4 females) working in these centers rated

students on the OM-GARS-2. And 7 teachers (4 males and 3 females) in these schools rated

students on the OM-GARS-2. All ratings were performed over a three-week period.

Methodology: To answer the research questions, two types of reliability indictors were

computed: (1) test-retest reliability, and (2) internal consistency reliability. Then, Pearson

correlation coefficient was computed between students’ scores on OM-ABC which is studied by

several researchers and the total score and the OM-GARS-2 subscales as well as total score.

Finally, a multivariate analysis of variance (MANOVA) was conducted where group (autistic vs.

normal) was set as an independent variable (factor) and the OM-GARS-2 subscales and Autism

index was set as criterion variable to test the hypothesis that there would be one or more mean

differences between groups.

Results: The results of the study showed the OM-GARS-2 had temporal stability and internal

consistency reliability. The OM-GARS-2 had criterion (type concurrent) validity and discriminant

validity. Conclusion: To conclude, the reliability and validity indices of OM-GARS-2 are very
 similar to that of the original GARS-2 (Gilliam, 2006) and other studies conducted internationally

6 Keywords: Gilliam Autism Rating Scale-2, Autism, Omani context.

7 1. INTRODUCTION

8 Autism is a disorder among a group of disorders under the umbrella of Autism Spectrum

9 Disorder (ASD) or Pervasive Developmental Disorder (PDD) (American Psychiatry Association, 1994,

10 2000). This disorder is noticed typically before the age of 3 and it has three defining core features: (a)

11 problems with social interactions, (b) impaired verbal and nonverbal communication, and (c) a pattern

12 of repetitive behavior with narrow, restricted interests (CDC, 2006). In the Diagnostic and Statistical

13 Manual of Mental Disorders – Fifth Edition, Text Revision (DSM-V-TR) these features were reduced

14 to two main characteristics; social communication and interaction and restricted, repetitive behavior

15 (Mazefsky, et al., 2013). According to the latest report published by the CDC, based upon the data

16 collected by the Autism and Developmental Disabilities Monitoring Network on 8-year-old children

17 living in 11 American provinces in 2010, about 1 in 68 children (or 14.7 per 1,000) were identified with

18 ASD. This new estimate is roughly 30% higher than the estimate for 2008 (1 in 88), roughly 60%

19 higher than the estimate for 2006 (1 in 110), and roughly 120% higher than the estimates for 2002

20 and 2000 (1 in 150) (CDC, 2010).

21 The growing rates of autism in recent years have led to considerable interest in its core

22 symptoms and diagnosis. Diagnosis is considered a fundamental and prerequisite step to initiate and

23 introduce special education services for children and adults with autism. Filipek, et al., (2000)

24 proposed that “the diagnosis of autism should include the use of a diagnostic instrument with at least

25 moderate sensitivity and good specificity for autism” (p.475). The authors advocated the Gilliam

26 Autism Rating Scale (GARS), among other measuring tools, as diagnosis tools of autism. A survey

27 published in 2008 found that 40% of school psychologists used the GARS-2 in the majority of their

28 ASD-related assessments (Allen, et al., 2008). The GARS-2 is a 42-item informant rating scale

29 designed to assist in the identification and diagnosis of autism and provide information on symptom
30 severity. The GARS-2 was built based on the definitions of autism that emerged from Diagnostic and

31 Statistical Manual of Mental Disorders- fourth edition, text revision (DSM-IV-TR) (American

32 Psychiatry Association, 2000) and the Autism Society of America. Gilliam (2006) stated that the

33 GARS-2 remains the only normed screening instrument based on these definitions. The GARS-2 was

34 normed using a sample of 1,107 individuals identified as diagnosed with autism and aged between 3

35 and 22 years.

36 The GARS-2 can be completed by parents, teachers and/or clinicians. Each of the items is

37 rated on a four-point frequency scale (i.e., 0 = Never Observed, 1 = Seldom Observed, 2 =

38 Sometimes Observed, and 3 = Frequently Observed). The 42 items are grouped to form three 14-

39 item subscales. (a) Social Interaction, (b) Communication, and (c) Stereotyped Behaviors. These

40 three scales are combined to create the overall Autism Index (AI). A parent interview is included

41 which taps into the child’s development during the first three years of life, however, item scores from

42 this interview are not factored into the overall AI. The Stereotyped Behaviors subscale focuses on

43 stereotyped behaviors, motility disorders and other unique and atypical behaviors. The

44 Communication subscale contains items that describe verbal and nonverbal behaviors that are

45 indicative of autism. Finally, the Social Interaction subscale defined the individual’s ability to relate

46 appropriately to people, events and objects (Gilliam, 2006).

47 For each of the GARS-2 subscales, the numeric responses from the 14 items are summed into

48 a total raw score. The total raw score is converted to a derived standard score (M = 10, SD = 3). The

49 sum of the standard scores from the three subscales is converted into the overall AI (M = 100, SD =

50 15), which is standardized to a deviation quotient metric. For non-communicative individuals, the

51 Communication subscale is omitted and the AI is calculated based on the other two subscales.

52 According to the manual, an AI score of 85 or higher indicates a “very likely” probability of autism,

53 scores between 70 and 84 suggest the probability of autism is “possibly,” and scores of 69 or below

54 indicate that the probability of autism is “unlikely” (Gilliam, 2006, pp. 31-32).
55 The GARS-2 manual reported reliability data for both internal consistency and stability (Gilliam,

56 2006). Internal consistency estimates were .88 for Social Interaction, .86 for Communication, .84 for

57 Stereotyped Behaviors, and .94 for the Autism Index. Corrected test-retest coefficients (1-week

58 interval) based on parent ratings of 37 children with autism were .88 for Social Interaction, .70 for

59 Communication, .90 for Stereotyped Behavior, and .88 for the overall Autism Index. The criterion-

60 related validity was established by computing correlation coefficients between the GARS-2 and the

61 Autism Behavior Checklist subscales (Krug, et al., 1993). The construct-identification validity was

62 established by examining (a) relationships of the GARS-2 subscales scores and age, (b) the internal

63 consistency of the GARS-2 subscales interrelationships, (c) the GARS-2 subscales standard scores

64 and Autism Index correlations, (d) evidence that the GARS-2 has practical value and ability to

65 differentiate autism from other groups (e.g., normal, mental retardation, and multiple disabilities).

66 Gilliam (2006) discussed several differences between the GARS and the GARS-2 including;

67 (1) the developmental disturbances subscale was revised and converted into an interview form to

68 allow examiners to evaluate the child’s development during early childhood. This procedures reduces

69 the time needed for completing the ratings, (2) some items were re-written clearly, (3) demographic

70 characteristics of the normative sample are keyed to the 2000 U.S. census, (4) all new norms were

71 created and the normative sample is more clearly described, (5) the total scores of the GARS-2 were

72 changed from Autism Quotient to Autism Index, (6) guidelines for interpreting subscales scores and

73 the Autism Index were changed, (7) a separate chapter is provided in which discrete target behaviors

74 for each item on the GARS-2 are defined and specific examples are given for applied behavior

75 analysis projects and other research purposes, and (8) a separate booklet “Instructional Objectives

76 for Children Who Have Autism” was developed to assist in the formulation of instructional goals and

77 objectives based on the results of the GARS-2 Tools.

78 The Omani Context

80 million native Arabic speakers. Omani population has no standard scale to diagnosis their children

81 who are at risk to be autistic, they have to go Jordan or Tunisia or Egypt to get a diagnosis.

82 Consequently, there appears to be a need for developing a tool that can help educators and clinician

83 to identify or diagnose the autistic children.

84 Versions of GARS-2

85 Li (2005) investigated whether a Chinese version and an English version of the GARS-2 were

86 measuring the same construct. The sample of the study included 20 bilingual Chinese-English

87 speaking parents who had at least one neurotypically developing child ages 2 years through 17

88 years, and who were immigrants in the United States. Scores on the two versions of the GARS-2

89 correlated highly and significantly for all subscales and for the Autism Index, suggesting that the

90 two versions are measuring the same construct. The subscales of the Chinese version of the

91 GARS-2 showed acceptable internal consistency. A serious limitation of Li’ study is the utilization

92 of a non-clinical sample.

93 Diken, Diken, Gilliam, Ardic, and Sweeney (2012) conducted a preliminary study to investigate the

94 validity and reliability of a Turkish Version of GARS-2 (TV-GARS-2). Participants included 436

95 children diagnosed with autism. Data were also collected from individuals diagnosed with

96 intellectual disability, with hearing impairment, and from typically developing children in order to

97 examine discrimination validity of the TV-GARS-2. Coefficient alpha of all subscales and the entire

98 instrument showed acceptable internal consistency. The test re-test reliability coefficients showed

99 acceptable temporal stability. The data provided several indices of TV-GARS-2 construct validity;

100 (1) non-significant correlation with students’ chronological age except for Stereotyped Behaviors (r

101 = .15, p <.01), (2) significant interrelationship among TV-GARS-2 subscales (r = .34 to .65, p <

102 .01), (3) item showed acceptable discriminating power, (4) significant corrected correlation

103 coefficients between the Autism index and the three subscales (r = .44 to .60, p < .01), and (5)
104 The TV-GARS-2 discriminated significantly among four groups of children; intellectual disability

105 group, hearing impairment group, normal development group, and autistic disorder group,

106 suggesting evidence of the TV-GARS-2 discriminant validity.

107 Al Jabery (2008) conducted a preliminary study to develop a Jordanian Arabic Version of the

108 Gilliam Autism Rating Scale (J-GARS-2). The sample included 100 students aged from three to 13

109 years and it was divided into two groups (50 students each): students with autism and students

110 with mental retardation. The test re-test reliability coefficients showed acceptable temporal

111 stability. Alpha coefficients and split half reliability showed acceptable internal consistency. The

112 total scores of the J-GARS-2 (Autism Index) correlated highly and significantly with the total score

113 on the Arabic version of the Autism Behavior Checklist (AV-ABC; Krug, Arick, & Almond, 1980;).

114 With the exception of the correlation between J-GARS-2 Stereotyped Behavior subscale and the

115 ABC Sensory subscale, all of the hypothesized correlations were found to be significant and

116 moderate to high in magnitude. The correlations between J-GARS-2 subscales raw scores and

117 age were not significant. All subscales raw scores have a strong correlation with the total score of

118 the J-GARS-2. All subscales raw scores have a strong correlation with each other except for the

119 correlation between the Stereotyped Behavior and the Communication subscales. The GARS-2

120 discriminated between students with autism group and students with mental retardation on all

121 subscales as of the J-GARS-2 and the Autism index.

122 Rationale and aims of the study

123 The growing rates of autism in recent years has led to considerable interest in its core

124 symptoms and diagnosis (CDC, 2010). Furthermore, the challenges faced with differential diagnoses

125 of autism, and the symptomatology of this disorder highlighted the need for assessment tools that

126 contribute to accurate diagnoses. Several measuring tools have been developed and used in

127 Western societies to screen and diagnose autism. However, we know little about the symptoms and

128 diagnosis of autism amongst native Arabs. For example, in Oman, the number of studies conducted
129 is limited and official statistics are apparently not available. One possible reason that Oman lags

130 behind in autism screening and diagnosis has, in part, been due to the lack of Arabic language

131 measures with acceptable psychometric properties and also to the fact that many Omanis do not

132 have an adequate command of the English language for the use of English language measures. As

133 such, there is a need for a valid and reliable autism diagnostic tool written in Omani and normed on

134 Omani-speaking respondents. Thus, the problem of this study emerged from the need to provide the

135 current tool practices of children and adults with autism in Oman with another valid and reliable

136 instrument to be utilized by professionals to enhance the diagnosis practices. Specifically, the present

137 study reports some initial findings about the psychometric properties of an Omani version of the

138 Gilliam Autism Rating Scale (OM-GARS-2) (Gilliam, 2006).

139 Questions of the study

140 The study intended to answer the following questions:

141 1. What are the correlation coefficients of test-retest and internal consistency reliability of the OM-

142 GARS-2?

143 2. What are the correlation coefficients between the OM-GARS-2 and the Omani version of

144 Autism Behavior Checklist (OM-ABC) in terms of subscales and total (the Autism Index for

145 OM-GARS-2 and the Total Sum for the OM-ABC) scores?

146 3. Does the OM-GARS-2 differentiate students labeled with Autism and normal students in terms

147 of each subscale score (stereotyped behaviors, communication, and social interaction) and the

148 Autism Index?

149 2. METHODS

150 The researchers used the descriptive approach, by distributing the Checklist to the sample.Sample

151 Autism sample. The autism sample included 45 children (25 males and 20 females) aged

152 between 8 and 14 years (M = 12.3, SD = .61) and enrolled in two public centers of autism care in two

153 governorates in Oman. These centers are supervised by the Ministry of Social Development. Children
154 in these centers are considered lower-functioning due to significant delays including cognitive, social,

155 and communicative impairments, which hinder them from attending classrooms within their respective

156 schools. Those children are diagnosed to suffer autism based on a clinical diagnosis made by a

157 licensed medical professional and/or psychologist in Oman or in another country.

158 Raters sample. A total of 8 teachers (4 males and 4 females) working in these centers rated

159 students on the OM-GARS-2. The number of years of teaching experience of those teachers ranged

160 from 2 to 7 years (M = 4.6, SD = .64). Most teacher raters had worked with the student being rated for

161 at least three months prior to the rating. Teaching staff raters were familiar with the general

162 characteristics of autism, as a result of their special education training or work experience. The

163 assessment process was created in order for each student to be rated by the staff member who knew

164 her/him best, while also maximizing the statistical independence of each case being rated. All ratings

165 were performed over a three-week period in the first semester of the school year 2015/2016.

166 Non-autism sample. The non-autism sample was selected as an available sample from the

167 schools which accepted to involve in the research. It included 45 children (23 males and 22 females)

168 aged 8-14 years (M = 12.6, SD = .47). The children were enrolled in two public schools in two

169 governorates in Oman. A total of 7 teachers (4 males and 3 females) in these schools rated students

170 on the OM-GARS-2. The number of years of teaching experience of those teachers ranged from 2 to

171 8 years (M = 4.9, SD = .51). Most teachers had worked with the students being rated for at least 4

172 months prior to the rating. Teaching staff raters were familiar with the general characteristics of ASDs,

173 as a result of their work experience and academic qualifications. The non-autism sample did not

174 suffer any difficulties.

175 The GARS-2

176 The GARS-2 is a 42-item behavioral checklist designed to identify persons with autism. The 42

177 items are grouped to form three 14-item subscales. (a) Social Interaction, (b) Communication, and (c)

178 Stereotyped Behaviors. These three subscales are combined to create the overall Autism Index (AI).
179 A parent interview is included which taps into the child’s development during the first three years of

180 life, but it is not part of the scoring system. The Stereotyped Behaviors subscale focuses on

181 Stereotyped Behaviors, motility disorders and other unique and atypical behaviors.The

182 Communication subscale contains items that describe verbal and nonverbal behaviors that are

183 indicative of autism. The Social Interaction subscale describes the individual’s ability to relate

184 appropriately to people, events and objects (Gilliam, 2006). All items of the GARS-2 can be rated on

185 a four-point frequency-based scale that ranged from 0 to 3 (i.e., 0 = Never Observed, 1 = Seldom

186 Observed, 2 = Sometimes Observed, and 3 = Frequently Observed). The GARS-2 can be completed

187 by parents, teachers, and/or clinicians.

188 Translation of the GARS-2

189 Two bilingual assistant professors of psychology and special education translated the GARS-2

190 from English into Omani Arabic using the back-translation method (OM-GARS-2). Two other bilingual

191 assistant professors of psychology and special education, working without referencing to the English

192 version of the GARS-2, independently translated the Arabic version back to English. Finally, one

193 certified translator and a bilingual professor of psychology and special education independently

194 compared the original English version of the GARS-2 with the new English version that was

195 translated back from Arabic, and rated the match between the two versions on a scale of 0 or 1. A

196 score of zero represented no match, whereas a score of 1 represented perfect match. The average

197 percentage of match was 96 % which could be considered highly acceptable (see, Maneesriwongul &

198 Dixon, 2004). Furthermore, interobserver agreement was calculated using SPSS Crosstabs function,

199 which produces a Kappa statistic for level of agreement. According to Cohen (1960), Kappa values

200 lie between ‐1.00 and 1.00, with zero indicating chance agreement, positive values indicating greater

201 than chance agreement, and negative values indicating less than chance agreement. Landis and

202 Koch (1977) categorized Kappa values from 0.41 to 0.60 as moderate and values above .60 as
203 substantial levels of agreement. The inter observer agreement Kappa value for the OM-GARS-2 was

204 .75.

205 Autism Behavior Checklist (ABC)

206 The ABC was published in 1980 (Krug, et al., 1980) and is part of a broader tool, the Autism

207 Screening Instrument for Educational Planning (ASIEP). The ABC is designed to be completed

208 independently by a parent or a teacher familiar with the child who then returns it to a trained

209 professional for scoring and interpretation. Although it is primarily designed to identify children with

210 autism within a population of school-age children with severe disabilities, the ABC has been used with

211 children as young as 3 years of age. The ABC has 57 items and each item is weighted according to

212 the degree to which the characteristic is a symptom of autism [1-is related in a small degree to 4-is

213 related in a strong degree]. For example, “whirls self for long periods of time” receives four points,

214 whereas “does not follow simple commands.” receives one point (Krug, et al., 1993). The items are

215 grouped into five scales: Sensory, Relating, Body and Object Use, Language, and Social and Self-

216 Help. The Total Score, which is the sum of all items in the five scales, is used as a fundamental

217 indicator of autistic disorder. A cut off score of 67 indicates a high likelihood of autism, a score below

218 53 indicates a low likelihood of autism, and a score between 53 and 67 indicates the need for more

219 investigations (Lord & Corsello, 2005). Al Hadramy (1995) developed the Arabic version of the ABC in

220 Oman (OM-ABC) using a sample of 114 children aged 10-12 years old. She reported that the ABC

221 has good internal reliability, and good sensitivity and specificity. Overall, studies indicated that the ABC

222 instrument has good psychometric properties to use in the Arabic region.

223 Procedures

224 The researchers of this study coordinated the data collection procedures as part of a two-year

225 research project by obtaining necessary official permissions and contacting the autism care centers

226 and public schools. Before starting data collection at the autism care centers, one of the researchers

227 in this study and a research assistant held a meeting with teachers in each center to explain the
228 purpose of the study and familiarize teachers with the instruments (OM-GARS-2 and OM-ABC). The

229 researcher and the research assistant emphasized the notion that participation in data collection is

230 voluntary and that collected data will be kept confidential and they will be used solely for research

231 purposes. Teachers were encouraged to read the OM-GARS-2 and the OM-ABC carefully before the

232 day of the meeting and they were given the chance to ask questions that were answered by the

233 researchers. This procedure intended to support the internal validity of the study by minimizing raters’

234 bias. Teachers were given two weeks to complete the OM-GARS-2 and OM-ABC and rate their

235 students. Teachers were blind to each other. They were instructed not to discuss students’ ratings

236 with each other to keep rating independency. The purpose of administering the OM-ABC was to

237 examine the criterion (type of concurrent) validity of the OM-GARS-2. Two weeks later, the teachers

238 were given and asked again to complete the OM-GARS-2 for their students (including pupils) for

239 purposes of examining the test-retest reliability of the OM-GARS-2. The data collection of the normal

240 sample followed the same procedures of the autism sample. Data collection took place during normal

241 classes at targeted schools. Teachers were given one week to complete the OM-GARS-2 and rate

242 their students. A research assistant individually collected the instruments from both the autism care

243 centers and the public schools and reviewed them to assure their full completion. Only students with

244 complete dataset (98%) were included in the analyses.

245 3. RESULTS

246 Question 1. What are the correlation coefficients of test-retest and internal consistency

247 reliability of the OM-GARS-2?

248 To answer this question, two types of reliability indictors were computed: (1) test-retest

249 reliability, and (2) internal consistency reliability.

250 For the test re-test reliability, Pearson correlation coefficient was computed for the autistic

251 students’ scores on the OM-GARS-2 over the two points of data collection. The results showed that

252 the correlation coefficients were .92 for Stereotype Behavior, .89 for Communication, .91 for Social
253 Interactions, and .93 for the entire instrument (Autism Index). All correlations were statistically

254 significant at .001. For internal consistency reliability, Cronbach’s alpha coefficients were computed

255 for the autistic students to judge the internal consistency of the OM-GARS-2 subscales as well as the

256 entire instrument (Autism Index). Results indicated that Alpha coefficients were .91 for Stereotype

257 Behavior, .90 for Communication, .87 for Social Interactions, and .89 for the entire instrument.

258 Question 2. What are the correlation coefficients between the OM-GARS-2 and the Omani

259 version of Autism Behavior Checklist (OM-ABC) in terms of subscales and total (the Autism

260 Index for OM-GARS-2 and the total sum for the OM-ABC) scores?

261 To answer this question, Pearson correlation coefficient was computed between students’

262 scores on OM-ABC and the total score and the OM-GARS-2 subscales as well as total score. Table 1

263 shows that all Pearson correlation coefficients were statistically significant.

264 Table 1
265 Pearson correlations coefficients between OM-GARS-2 subscales raw scores and total score (Autism
266 Index) and the OM-ABC subscales raw scores and total sum
267

OM-ABC subscales

Sensory Relatin Body and languag Social and ABC

OM-GARS-2 g object use e Self-help Sum

subscales

Stereotype Behavior .52** .42** .39** .33* .40** .56**

Communication .49** .50** .48** .45** .32* .60**

Social Interactions .44** .43** .28* .31* .29* .57**

Total (Autism Index) .47** .47** .50** .41** .37* .52**

268 Note. N = 45. **p < .01. *p < .05

269
270 Question 3. Does the OM-GARS-2 differentiate students labeled with Autism and normal

271 students in terms of each subscale score (stereotyped behaviors, communication, and social

272 interaction) and the total score (Autism Index)?

273 To answer this question, a multivariate analysis of variance (MANOVA) was conducted where group

274 (autistic vs. normal) was set as an independent variable (factor) and the OM-GARS-2 subscales and

275 Autism index was set as criterion variable to test the hypothesis that there would be one or more

276 mean differences between the groups. A statistically significant MANOVA effect was obtained, Pillais’

277 Trace = .43, F (4, 40) = 16.94, p < .001. The multivariate effect size was estimated at .23, which

278 implies that 23% of the variance in the canonically derived dependent variable was accounted for by

279 group factor. A series of one-way ANOVA’s on each of the four dependent variables was conducted

280 as a follow-up tests to the MANOVA. Univariate results demonstrated a significant effect (p < .01) for

281 Stereotype Behaviour, F (1, 43) = 7.33, partial η2 = .10, Social Interaction, F (1, 43) = 4.30, partial η2 =

282 .09, and Communication, F(1, 43) = 11.20 partial η2 = .13. Partial η2 can vary in magnitude with <

283 .01indicting small effect size, >.02 to < .06 indicating medium effect size, and > .07 to >.14 indicating

284 large effect size. Table 2 show mean differences between the autistic group and the normal group.

285 These differences are presented pictorially in Figure 1.

286 Table 2
287 Mean differences between autistic group and normal groups in three subscales of the OM-GARS-2
288 and the Autism index.
289
Group/Factor Stereotype Communication Social Interaction Autism Index

Behaviour

Autistic 3.3 3.5 3.6 3.5

Normal 2.2 2.4 2.2 2.4

Mean 1.1 1.1 1.4 1.1

differences*

Cohen’s D .53 .53 .64 .65

291

292

293

294 Figure 1. Mean differences between autistic group and normal groups in three subscales of the OM-
295 GARS-2 and the total score (Autism index)
296
297 4. DISCUSSION

298 The present study reports some initial findings concerning the psychometric properties (validity

299 and reliability) of the Omani version of the Gilliam Autism Rating Scale-2 (OM-GARS-2) (2006) within

300 an Omani context. The first step in this study was to translate the GARS-2 from English into Arabic.

301 The goal was to develop an Arabic translated version of the GARS-2 that typically matches the

302 original English version of the scale and that is culturally appropriate for the Omani context. According

303 to the Center for Aging in Diverse Communities, Measurement and Methods Core (2007) “A well-

304 translated survey instrument should have semantic equivalence across languages, conceptual

305 equivalence across cultures, and normative equivalence to the source survey. Semantic equivalence

306 refers to the words and sentence structure in the translated text expressing the same meaning as the

307 source language. Conceptual equivalence is when the concept being measured is the same across

308 groups, although wording to describe it may be different. Normative equivalence describes the ability

309 of the translated text to address social norms that may differ across cultures.” (2007, p.1). The
310 present study employed the back-translation strategy. A back translation was conducted by an

311 independent translator who has had no previous exposure to the document being translated. Again,

312 the emphasis of the back translation is the conceptual and cultural equivalence (Abd-El-Fattah, et al.,

313 2014). The present study calculated the percentage of agreement of two raters concerning the match

314 between the two versions of the GARS-2; the original English version and the English version that

315 was translated back from Arabic. Although the percentage of agreement was substantially high

316 (96%), the researcher preferred to calculate Kappa statistics for interobserver agreement because

317 percentage agreement does not correct for chance agreement (Abd-El-Fattah, 2013). The kappa

318 statistics was .75 suggested substantial interobserver agreement which implied that the two versions

319 of the GARS-2 were matched. This finding supported the translation of the GARS-2 into Arabic.

320 The second step was to obtain the psychometric properties for Omani version of the GARS-2

321 (OM-GARS-2) including validity and reliability indicators to support the entire instrument. The first

322 question of this study concerned the reliability of the OM-GARS2. Reliability indicators were

323 calculated by two methods; the test re-test reliability and the internal consistency reliability. Results

324 indicated strong reliability indicators for the OM-GARS-2 that would support its consistency in

325 measuring the same concept, that is, autism. An examination of the results revealed that the values

326 of the test-retest reliability were high (.92, .89, .91, and .93). One can conclude that the translated

327 version of the GARS-2 has strong temporal stability in measuring the autism disorder. This finding

328 implies that teachers were highly consistent in rating their students during the two times of

329 administration of the OM-GARS-2. This result confirms that teachers understood the OM-GARS-2

330 items and indicate similar rates. This may imply that teachers as raters are highly educated and well

331 qualified to deal with a rating scale of autism. Moreover, the reliability coefficient of the internal

332 consistency (Cronbach Alpha) were also high to indicate a strong internal consistency. These

333 findings are very similar to that of the original GARS-2 (Gilliam, 2006) and other studies conducted
334 internationally in Jordan, (Al Jabery, 2008; Alsqour, 2014), Greece (Tafiadis, et al., 2008), Turkey

335 (Diken et al., 2008), and China (Li, 2005).

336 The second question of the present study concerned the validity of the OM-GARS-2. The

337 criterion (concurrent type validity) of the OM-GRAS-2 was established by calculating Pearson

338 correlation coefficient between the OM-GARS-2 and the OM-ABC, hypothesizing that the two

339 instruments are measuring the same construct, that is, autism disorder. The computed correlations

340 coefficients were all statistically significant, suggesting significant correlations between both

341 instruments. One important point to note when examining these correlations is that these correlations

342 were moderate to high in magnitude and that could be attributed either to the small size of sample in

343 which using a bigger sample might improve the correlations, or to raters understanding of the

344 behavioral manifestations of their students’ autistic behaviors and their abilities to rate the core

345 deficits as measured by different sample of items (Al Jabery, 2008).

346 The third question of the present study concerned the discriminant validity of the OM-GARS-2.

347 A multivariate analysis followed by several univariate analyses and post-hoc analyses (Least

348 Significance Difference “LCD”) for pairwise comparison showed that the autistic group scored higher

349 than the normal group on the OM-GARS-2 Stereotype Behavior Communication, Social Interaction,

350 and total score. These findings indicated that OM-GARS-2, can be used in differentiating persons

351 with autism from other persons with normal development.

352 Ethical approval:

353 We take acceptance from the director of the centres, and the ministry of Education

354 Limitations and suggestions

355 Because this research used only 90 subjects (45 autism sample and 45 non-autism sample) which

356 were not enough to report the validity and reliability of the instruments. This should be done in further

357 research with more subjects.

359 result only comes from are teachers who have expertise and are familiar with observing and

360 evaluating behavioral problems. The reliability and validity of this instrument for the families should be

361 investigated.

362 CONCLUSION

363 To conclude, these results offer a valid and reliable tool for autistic children in Oman, to be

364 diagnosis in their country. It help clinician and specialist to have the opportunity to early identification

365 and then, early intervention. Findings of this study also highlights the GARS-2 as a culturally robust

366 scale.

367 COMPETING INTERESTS

368 Authors declared that no competing interests exist.

369 Ethical: NA

370 Consent: NA

371

372 REFERENCES

373 1- Abd-El-Fattah, S. M. A cross-cultural examination of the Aggression Questionnaire Short form

374 among Egyptian and Omani adolescents. Journal of Personality Assessment. (2013); 95: 539-

375 548.

376 2- Abd-El-Fattah, S. M., Al Sinani, Y., El Shourbagi, S., Fakhroo, H. A. Using Rasch analysis to

377 examine the dimensionality structure and differential item functioning of the Arabic version of

378 the perceived Physical Ability Scale for Children. Australian Journal of Educational &

379 Developmental Psychology. (2014); 14: 29-44

380 3- Al Hadramy, M. S. The psychometric properties of the Autism Behavior Checklist (ABC) within

381 an Omani context. Unpublished Master Thesis, Alexandria University, Egypt; (1995).
382 4- Al Jabery, M. A. The examination of validity and reliability indicators of the Jordanian translated

383 Arabic version of the Gilliam Autism Rating Scale (GARS-2). Unpublished doctoral

384 dissertation, Wayne State University, Michigan; (2008).

385 5- Allen, R. A., Robins, D. L. & Decker, S. L. Autism spectrum disorders: Neurobiology and

386 current assessment practices. Psychology in the Schools, (2008); 45(10): 905-917

387 6- Alsqour, I.A. Derivation of performance norms for the Jordanian Arabic version of Gilliam

388 Autism Spectrum Disorder rating scakle for ages 3 to 13 years (Second Edition), Unpublished

389 thesis, Univeristy of Jordon, Amman; (2014).

390 7- American Psychiatry Association. Diagnostic and statistical manual of mental disorders (4th

391 ed). Washington, DC: American Psychiatry Association; (1994).

392 8- American Psychiatry Association. Diagnostic and statistical manual of mental disorers (4th ed.

393 Rev). Washington, DC: American Psychiatry Association; (2000).

394 9- Center for Aging in Diverse Communities. Guidelines for translating surveys in cross cultural

395 research. (2007). Accessed 14 March 2017.

396 Available: http://www.rcmar.ucla.edu/measurement_surveys.php

397 10- Centers for Disease Control and Prevention. Parental report of diagnosed autism in children

398 aged 4–17 years—United States, 2003–2004. MMWR. (2006); 55(17): 481-486, Accessed

399 14 December 2016.

400 Available: http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5517a3.htm

401 11- Speech & Hearing Associates. (2010). 10 things you need to know about CDC's latest report

402 from the Autism and Developmental Disabilities Monitoring Network. Accessed 29 January

403 2016.

404 Available: http://www.cdc.gov/features/dsautismdata/

405 12- Cohen, J. A coefficient of agreement for nominal scales. Educational and Psychological

406 Measurement. (1960); 20: 37‐46.

408 version of Gilliam Autism Rating Scale-2. Education and Science. (2012); 37(166): 318-328.

409 14- Filipek, P. A., Accardo, P. J., Ashwal, S., Baranek, G. T., Cook, E. H., Dawson, G., Volkmar, F.

410 R. Practice parameter: Screening and diagnosis of autism. Report of the quality standards

411 subcommittee of the American Academy of Neurology and the Child Neurology Society.

412 Neurology. (2000); 55(4): 468-479.

413 15- Gilliam, J. E. Gilliam Autism Rating Scale – Second Edition. Austin, TX: Pro-Ed; (2006).

414 16- Krug, D. A., Arick, J. R., & Almond, P. J. Autism screening instrument for educational

415 planning. Austin, TX: PRO-ED; (1980).

416 17- Krug, D. A., Arick, J. R., & Almond, P. J. Autism Screening Instrument for Educational Planning

417 (ASIPE2) (2nd ed.). Austin, TX: PRO-ED; (1993).

418 18- Landis, J. R., & Koch, G. G. An application of hierarchical kappa‐type statistics in the

419 assessment of majority agreement among multiple observers. Biometrics. (1977); 33: 363‐374.

420 19- Li, N. (2005). Preliminary validation of the Childhood Autism Rating Scale – Second Edition

421 Questionnaire for Parents or Caregivers (CARS2-QPC) and the Gilliam Autism RatingScale

422 (GARS-2) with a Chinese-speaking population. Unpublished Master Thesis. University of

423 Eastern Kentucky

424 20- Lord, C., & Corsello, C. (2005). Diagnostic instruments in the autism spectrum disorders. In

425 F.R. Volkmar, R. Paul, A. klin, & D. J., Cohen (Eds.), Handbook of Autism and Pervasive

426 Developmental Disorders (3rd ed., pp. 730–772). New York: John Wiley & Sons, Inc.

427 21- Maneesriwongul, W., & Dixon, J. K. Instrument translation process: A methods review. Journal

428 of Advanced Nursing Research. (2004); 48: 175-186.

429 22- Mazefsky, C. A., McPartland, J. C., Gastgeb, R. Z., & Minshew, N. J. Brief report:

430 Comparability of DSM-IV and DSM-5 ASD research samples. Journal of Autism and

431 Developmental Disorders. (2013); 43: 1236-1242.

433 (GARS - 2), a pilot study for the Greek autistic population. Poster presented at

434 International Society on Brain and Behavior: 3rd International Congress on Brain and Behavior

435 (28 November – 2 December), Thessaloniki, Greece.

436