The participants were recruited through the online survey approach via convenience sampling. They were students/staff from three universities in Guangdong province, China. The online survey package consisted of a cover page concerning information such as purpose of the research, demographic information items (i.e. age and gender) and measurements used in the study. In total, 373 individuals voluntarily participated in the present research and completed the survey. The total sample was randomly split into two parts as calibration and validations samples for subsequent statistical analysis. The demographic information can be seen in Table 1.
Variable Total Sample (N = 373) Calibration Sample (N = 187) Validation Sample (N = 186) Age (years) Mean ± SD 21.41 ± 3.36 21.19 ± 2.87 21.62 ± 3.79 Range 18-42 18-39 18-42 Gender, n(%) Male 207 (55.50%) 101 (54.01%) 106 (56.99%) Female 166 (44.50%) 86 (45.99%) 80 (43.01%) Note. SD, Standard deviation.
Table 1. Demographic Statistics of Participants
Prior to questionnaire administration, ethical clearance was obtained from the local university research ethics committee. Translation and back translation techniques[32, 33] were adopted in the present study, to ensure equivalency when translating the NSS-SF into Chinese (CNSS-SF). Specifically, two professional translators were invited to translate the original English version into Chinese, independently. A comparison was made between their translations, and modifications were made until consensus was achieved between the two translators. Subsequently, two other professional translators were invited to translate the CNSS-SF back into English. Comparison was conducted to locate gaps to the original English version, and revisions were made until consensus was reached between the two translators. Subsequently, five Chinese adults were invited to check the comprehensibility of the CNSS-SF. Based on their assessment, the translated scale was well comprehensible and no suggestion was elicited for further adjustment. Hence, the CNSS-SF for subsequent administration was finalized.
The anonymous, structured questionnaire package was created and posted online via an online survey platform (https://www.wjx.cn/), which is popular in China. Invitation pamphlets/posters containing basic information of the research as well as online survey quick response (QR) code were created. Pamphlets were circulated in universities and posters were posted on poster board on campuses by the researcher. Students/staff interested in the present investigation could access the electronic questionnaire battery via QR code. No monetary/material incentive was provided for participants. In the cover page of the online survey, participants were informed the aim of the study, and their participation was anonymous, confidential and voluntary. They were entitled to withdraw at any time. Furthermore, they were asked to respond honestly, as there were neither right, nor wrong answers.
Chinese Version of the Weinstein Noise Sensitivity Scale-short Form (CNSS-SF) The scale consists of 5 items, deriving from the long-form scale, originally coded as items 7, 8, 18, 19, and 21. The six-point Likert scale is applied, ranging from 'strongly disagree (1)' to 'strongly agree (6)'. Among the 5 items, one item (item 2) is negatively worded. The final score is obtained by reversely coding item that is negatively worded and summing up all the items. A higher score indicates higher degree of noise sensitivity.
Chinese Version of the State-trait Anxiety Inventory Form Y-2 (STAI-C, Y-2) The STAI, as an instrument to assess trait and state anxiety, has been widely used. It comprises a total of 40 items, with 20 items for state anxiety (Form Y-1) and 20 items for trait anxiety (Form Y-2). The four-point Likert scale is used for quantification, ranging from 1 (almost never) to 4 (almost always). In the present study only the trait anxiety sub-scale (Form Y-2) was adopted, which assesses stable manifestation of anxiety by means of asking respondents how they typically feel. Particularly, items 21, 23, 26, 27, 30, 33, 34, 36, and 39 in the Form Y-2 are negatively worded. The scoring of the scale is done by reversely coding items that are negatively worded and summing up all the items. A higher resulting score denotes a higher level of trait anxiety. Since its development[34, 35], the STAI has been extensively translated and well-validated in different languages, including Greek, Malaysian, Japanese, Portuguese, Spanish, Norwegian, French, Brazilian, and Chinese, revealing adequate psychometric property.
Descriptive statistical analysis was adopted for mean value, standard deviation, skewness and kurtosis of each item. The data was evenly split into two samples, in a random fashion. Specifically, sample 1 (N = 187) served as the calibration sample, while sample 2 (N = 186) served as the validation sample. CFA was performed to examine the factorial validity of the CNSS-SF in AMOS 22.0. The sample size met the ratio of 10:1 required in CFA between sample size and total items of the scale. Maximum-likelihood (ML) estimation with 5, 000 bootstrap samples was introduced to examine the factorial validity of the CNSS-SF. As for model fit indices used to examine the fit of the constructed model, the ratio of χ2/df with a value of less than 3 indicates reasonable fit[48, 49]. Meanwhile, in line with the recommendations of Hu and Bentler, comparative fit index (CFI), Tucker-Lewis index (TLI), standardized root mean square residual (SRMR) and the root mean square error of approximation (RMSEA) accompanied by its 90% confidence interval (90% CI) were also employed in the present study. Cut-off values of 0.90 and 0.95 for CFI/TLI demonstrate acceptable and good model fit, respectively. Likewise, cut-off values of 0.08 and 0.06 for RMSEA/SRMR represent acceptable and good fit, respectively. Regarding factor loading determination, the standardized factor loading cut-off larger than 0.40 was applied, which has been widely used in past researches adopting factor analysis[51, 52].
Measurement invariance (MI) was introduced to examine whether the scale would exhibit gender invariance. In particular, a configural model and two increasingly constrained models, viz. measurement weights and structural covariances models, were assessed. Since error variance covariance is of little interest and usually deemed unnecessary, it was not performed in the present study. A non-significant χ2 value, an alteration of CFI of less than 0.01 between competing models, as well as an alteration of RMSEA of less than 0.015 between comparison models, served as benchmark of group invariance assessment[53, 54].
Nomological validity was tested for the CNSS-SF by examining its association with the STAI-C-Y2, which was found linked to noise sensitivity in previous studies[23, 25]. In line with previous results, it was hypothesized that respondents' scores in the CNSS-SF would be positively correlated with scores of trait anxiety.
Estimation of internal consistency was carried out using Cronbach's alpha and composite reliability. The advantage of adopting composite reliability is that it is able to account for measurement errors of all indicators, as well as provide better estimation of the internal consistency reliability of the measure. A value of 0.70 or greater is considered evidence of acceptable internal consistency of the CNSS-SF scale for both Cronbach's alpha[57, 58] and composite reliability.
The descriptive analysis results for both the calibration sample and the validation sample are displayed in Table 2. Sufficient model fit for both the calibration and validation sample data was demonstrated in CFA analyses. The specific model fit statistics and standardized factor loadings for the two sample data are displayed in Table 3 and Table 4, separately.
Item Calibration Sample (N = 187) Validation Sample (N = 186) Mean SD SK KU Mean SD SK KU 1 4.37 1.256 -0.530 -0.264 4.37 1.241 -0.434 -0.285 2 4.02 1.222 -0.238 -0.462 3.97 1.311 -0.289 -0.346 3 3.99 1.416 -0.163 -0.797 4.05 1.421 -0.153 -0.927 4 4.56 1.240 -0.403 -0.820 4.46 1.395 -0.579 -0.672 5 4.19 1.146 -0.103 -0.856 3.89 1.249 -0.031 -0.632 Note. SD, Standard deviation; SK, Skewness; KU, Kurtosis.
Table 2. Item Means, Standard Deviation, Skewness and Kurtosis across Samples
Model χ2 df χ2/df P CFI TLI RMSEA (90% CI) SRMR Calibration sample (N = 187) 9.207 5 1.841 0.101 0.972 0.944 0.067 (0.000-0.135) 0.0377 Validation sample (N = 186) 10.353 5 2.071 0.066 0.969 0.939 0.076 (0.000-0.142) 0.0366
Table 3. Model Goodness-of-fit for Calibration and Validation Samples
Item Content Calibration Sample
(N = 187)
(N = 186)
1. I get annoyed when my neighbors are noisy. 0.655 0.406 2. I get used to most noises without much difficulty. 0.552 0.449 3. I find it hard to relax in a place that's noisy. 0.531 0.529 4. I get mad at people who make noise that keeps me from falling asleep or getting work done. 0.587 0.765 5. I am sensitive to noise. 0.584 0.759 Note. The Chinese version of the NSS-SF is available on request if necessary. Please direct your request to the correspondence author in Chinese or English.
Table 4. Standardized Factor Loadings across Samples
Nomological validity is regarded as a core facet for assessing the overall validity of a measurement. It was tested by investigating the association between score of CNSS-SF and trait anxiety. As expected, score of noise sensitivity was significantly positively correlated with score of trait anxiety (r = 0.155, P < 0.01).
The reliability of the CNSS-SF was evaluated by internal consistency (Cronbach's alpha and composite reliability) in the calibration sample and the validation sample, respectively. For the calibration sample, Cronbach's alpha and composite reliability were 0.715 and 0.720, respectively, while in the validation sample, Cronbach's alpha and composite reliability were 0.712 and 0.726, respectively. Internal consistency values in both samples exceeded the acceptable standard of 0.7, demonstrating that the CNSS-SF had acceptable reliability.
Results showed that the unconstrained model (default model) specified for male and female participants demonstrated sufficient goodness-of-fit. When the factor loadings were further constrained to be equal across gender, the model displayed adequate goodness-of-fit, which supported the factor loadings invariance across gender. In the third model, an additional constraint (set the factor variances, as well as co-variances, to be equal across males and females) was added, whilst the model still maintained acceptable goodness-of-fit overall.
Table 5 displays goodness-of-fit indices for the invariance analysis of the CNSS-SF. It revealed that the χ2 difference between M1 and M2 was insignificant (P = 0.225), and there was no substantial change in the CFI value (ΔCFI = 0.006 < 0.01), as well as in the RMSEA value (ΔRMSEA = 0.001 < 0.015). Therefore, we concluded that the factor loadings of CNSS-SF model were invariant across gender. Likewise, results exhibited that the χ2 difference between M2 and M3 was insignificant (P = 0.240), and there was no substantial change in either the CFI value (ΔCFI = 0.006 < 0.01), or the RMSEA value (ΔRMSEA = 0.000 < 0.015), suggesting invariance in factor variances/co-variances. In sum, the above results suggested that factor loadings, factor variances, as well as co-variances, were invariant across gender (male and female participants), in the current study.
Model Model Comparison χ2 df P CFI ΔCFI TLI RMSEA
ΔRMSEA SRMR Male - 7.205 5 0.206 0.988 - 0.976 0.046
- 0.0305 Female - 6.039 5 0.302 0.992 - 0.984 0.035
- 0.0328 M1 - 13.244 10 0.210 0.990 - 0.979 0.030
- 0.0305 M2 M1 vs. M2 18.920 14 0.168 0.984 0.006 0.977 0.031
0.001 0.0382 M3 M1 vs. M3 19.993 15 0.172 0.984 0.006 0.978 0.030
0.00 0.0378 Note. M1, Unconstrained model; M2, Equality of factor loading; M3, Equality of factor loading, factor variances/co-variances.
Table 5. Measurement Invariance across Gender (male = 207; female = 166)