真分数模型用于潜变量测评的弊端和效度改进的建议——以顾客满意度测量为例

The Defect of Latent Variable Measurement by Using True Score Model and Reliability Improvement——Taking Customer Satisfaction Measurement as an Example

顾客满意度测量与心理测量不同,与心理测量关注个体差异相比,顾客满意度测量更关注不同顾客反映出对某产品或者服务的共同评价。因顾客在评价产品和服务时会受某些共同的稳定的心理特征(比如城乡差异)的影响,用心理测量中的真分数模型对顾客满意度进行测量会降低测量的效度。通过分解随机误差项的思路对真分数模型进行了修正,给出了具体的操作程序,并通过实证研究对结果进行了验证。顾客满意度测量是潜变量测量的典型代表,修正的真分数模型同样可以运用于企业声誉、团队安全感等其他潜变量的测评。

Compared to psychological measurement considering individual difference,customer satisfaction measurement reflects the common evaluation of a product or service.Because customer＇s evaluation will be affected by some common stable psychological characteristics（such as urban-rural differences）,using the current measurement of customer satisfaction measurement model will reduce the validity of the measurement.By decomposing random error,the true score model can improve the validity.Therefore,this study gives a detailed operating procedures of using true score model,and through empirical research,the corrected true score model can also be applied to corporate reputation,team security assessment and other latent variables.