- 期刊
- OpenAccess
不同等级数利克特量表的比较研究－以满意度研究为例
Comparative Study on Variant Degree Likert Scale on Satisfaction Research
摘要
在市场研究中不同等级数的利克特量表都在广泛应用，但是缺少不同等级的利克特量表在调查中的适用性的系统研究。本研究以在线调查和电话访问两种访问为基础，系统地研究和比较了满意度研究中的5、6、7、10等级的利克特量表的结果、效度与信度、鉴别力以及选项分布特征发现：不同利克特量表的所测量的结果尽管分数差异较大但从结果中所得结论基本一致。但量表等级越多，鉴别力就越高，同样样本下可以得到更准确的结果或者反过来可以节约样本量，在CATI访问中，等级越多的量表所需要的样本量越少。十点量表所需要的样本量仅为五点量表的54%，七点量表所需样本量为五点量表的74.5%，能够节约大量的访问成本。而在在线调查中，六点量表、七点量表与十点量表均能够节约17.5%-28.5%的样本量。所以在能够看到选项的访问中，如在线调查或者面访，适用10点量表或7点量表；在不能够看到选项的访问中，主要指CATI，适用10点量表或者5点量表。对10点量表采用Opion-keyed修正，将1-4点统一赋分为4将大大提高量表的正态性从而有助于其他统计分析，而不改变量表结果、不降低量表的鉴别力。
並列摘要
The variant Likert degree scale is widely used in the market research. However, there are not many researches on the applicability of different degree scales. This study systematically compared the result, effect, validity, discrimination and distributing character of 5/6/7/10 degree Likert satisfaction scales through two research methods-online and telephone interviews. The result of this research indicates: the conclusions from different degrees Likert scales are the same, even though there is difference in the numeric result in each scale. Scales with more degrees can reflect more detailed results. Given the same sample size, we can obtain more precise results, or in another words, we can save sample size to obtain equal results. In CATI/telephone interview, scales with more degrees need smaller sample size. The required sample size of 10-degree scale is 54% of that of 7-degree scale, and the required sample size of 7-degree scale is 74.5% of that of 5-degree scales. Scale with more degrees will save a lot of interview cost. In online interview, 6-degree, 7-degree and 10-degree scales all can save approximately 17.5%-28.5% of the sample size. Therefore, when the multiple choice answers are visible by the interviewee, such as on-line or face-to-face interviews, 10-degree or 7-degree scale is more convenient, while when the multiple choice answers are invisible, such as CATI/telephone interview, 10-degree or 5-degree scales are more convenient. If results from 1 to 4 point are all replaced by 4 point through the Option-keyed technique, the normalization will improve greatly and this will be helpful for other statistical analysis without changing the result and its detailed precision.