網路問卷與傳統問卷之比較：多樣本均等性方法學之應用

Comparison of Internet and Paper-and-pencil Questionnaires in Taiwan Using Multi-sample Invariance Approach

游森期(Sen-Chi Yu) ； 余民寧(Min-Ning Yu)

《測驗學刊》 53卷1期 (2006/06) Pp. 103-127

https://doi.org/10.7108/PT.200606.0103

摘要

This study examines whether Internet questionnaires are as psychometrically invariant as traditional mailed questionnaires. A random sample of 2400 teachers in Taiwan was divided into experimental and control groups. The experimental group was asked to complete the electronic version of the CES-D (Center for Epidemiologic Studies Depression Scale) placed on the internet, whereas the control group was invited to complete the traditional paper-based CES-D.
The multi-sample invariance approach derived from structural equation modeling (SEM) was applied to analyze the collected data. The analytical results show that the two groups have equivalent factor structures. That is, the two groups exhibit equal factor loadings, covariance matrix for latent factors and measurement errors. The results also show that the items in CES-D function equivalently in the two groups. That is, the instrument exhibits equal validity and reliability when administrated to the two groups. Moreover, the cross validation was also approved by the results. These findings demonstrate that the Internet questionnaires and traditional mailed questionnaires had equal factor structures.
Given that the factor structures were found to be equivalent, the equal latent mean test was then performed. The analytical results demonstrate that the four latent means in CESD are riot totally equal between two groups. Concerning the four latent factors in CES-D, the latent means of ”depressed mood”, ”positive affect” and ”interpersonal problems” are not significantly different between these two groups. However, the difference between the ”somatic symptoms” latent means of these two groups is statistically significant at a=.01. To investigate the effect size of such difference, Cohen d statistics was utilized. However, Cohen d statistics was only .172, indicating a small effect size. That is, the minor difference in somatic symptoms, 0.19 point, albeit statistically significant, does not apparently lead to meaningful clinical significances (effect size) in practice.
Analytical results indicate that Internet questionnaires and traditional mailed questionnaires have equal factor structures to. Both CES-D questionnaires exhibit equal reliability, validity and factor structure, and similar latent means. In conclusion, internet questionnaires represent a promising alternative to traditional paper-based questionnaires.

關鍵字

網路問卷 ； 憂鬱 ； 結構方程式模式 ； 多樣本均等性

並列摘要

This study examines whether Internet questionnaires are as psychometrically invariant as traditional mailed questionnaires. A random sample of 2400 teachers in Taiwan was divided into experimental and control groups. The experimental group was asked to complete the electronic version of the CES-D (Center for Epidemiologic Studies Depression Scale) placed on the internet, whereas the control group was invited to complete the traditional paper-based CES-D.
The multi-sample invariance approach derived from structural equation modeling (SEM) was applied to analyze the collected data. The analytical results show that the two groups have equivalent factor structures. That is, the two groups exhibit equal factor loadings, covariance matrix for latent factors and measurement errors. The results also show that the items in CES-D function equivalently in the two groups. That is, the instrument exhibits equal validity and reliability when administrated to the two groups. Moreover, the cross validation was also approved by the results. These findings demonstrate that the Internet questionnaires and traditional mailed questionnaires had equal factor structures.
Given that the factor structures were found to be equivalent, the equal latent mean test was then performed. The analytical results demonstrate that the four latent means in CESD are riot totally equal between two groups. Concerning the four latent factors in CES-D, the latent means of ”depressed mood”, ”positive affect” and ”interpersonal problems” are not significantly different between these two groups. However, the difference between the ”somatic symptoms” latent means of these two groups is statistically significant at a=.01. To investigate the effect size of such difference, Cohen d statistics was utilized. However, Cohen d statistics was only .172, indicating a small effect size. That is, the minor difference in somatic symptoms, 0.19 point, albeit statistically significant, does not apparently lead to meaningful clinical significances (effect size) in practice.
Analytical results indicate that Internet questionnaires and traditional mailed questionnaires have equal factor structures to. Both CES-D questionnaires exhibit equal reliability, validity and factor structure, and similar latent means. In conclusion, internet questionnaires represent a promising alternative to traditional paper-based questionnaires.

並列關鍵字

Internet questionnaire ； depression ； structural equation modeling ； multi-sample invariance

參考文獻

張德勝(2005)。「學生評鑑教師教學」信效度之探討：紙筆和網路調查之比較。中華心理學刊。47(2)，113-125。

Alreck, R. L.(1995).The survey research handbook.Chicago:Irwin.

American Psychiatric Association(1994).Diagnostic and statistical manual of mental disorders (Fourth Edition).Washington D.C:American Psychiatric Association.

Birnbaum, M. H.(2004).Human research and data collection via the Internet.Annual Review of Psychology.55,803-822.

Bollen, K.A.(1989).Structural equations with latent variables.New York:John Wiley & Sons.