Constructing Self-esteem Scale Under The Assumption Of An Ideal Point Response Process

Self-esteem is that one's evaluating about the degree of deserving respect or the importance. Self-esteem is the core of personality, and the important field of psychological research. At present, self-esteem scale is constructed by the method of Likert, which have some problems. When analyzing and filtering items, we must follow the policies of high correlation of item-total score and factor loading. The method just contains the items which locate in two terminals of trait, and delete the items which locate in middle of trait. If in this case, the test can distinguish the tester is high level or low, which hasn't precise measurement in total interval of trait. The second, some items have to reverse scoring, whose logistical foundation is not strict. The third, item difficulty statistics computed by the proportion of Affirmative Response have no direct correspondence to item content.The reason maybe that the method follows the method of cognitive ability scales, but cognitive ability scale is a optimum behavior scale, and personality and self-esteem scale is typical behavior scale. Both methods have different response process. Cognitive ability scales have a dominance response process: if the tester and item are located in trait field, when the tester's location is more aggressive than item, the tester will make affirmative answer. the shape of the item response function is monotonic increasing. But personality scale have an ideal point process: if the tester and item are located in trait field, when the tester's location is approaching the item, the tester will make affirmative answer; when the distance between the tester and item, the tester will make negative answer, whether the location of tester is higher or lower than item. The shape of item response function is a bell.Some researchers compare the model-data fit of personality scale and two models (ideal point model and dominance model), found that the ideal point model has well fit and more precise measurement.Chernyshenko and Stark have constructed responsibility scale under the assumption of ideal point process. This study tries to constructing self-esteem scale under the assumption of ideal point process. The method is:1.writing items, experts evaluating, the first testing, and forming the pretest scale;2.merging the pretest scale and RSES(Rosberg Self-Esteem Scale),and measuring 1182 undergraduate;3.making use of GGUM2004 and estimating the parameter of items and testee;4.factor analysis, making use of MODFIT and computing the Chi-square fit statistics for pairs and triplets of items, and delete the items which are not fit;5.forming the formal scale,anaysing the reliability and validity;6.comparing the formal scale with RSES.The results of study found that the formal test constructed under assumption of ideal point process have some advantages:1.form comparing of test information function, we found that the formal scale have high reliability, and have precise measurement at every level.2.the tester's score(theta) estimated by GGUM2004 have Significant correlation with RSES(0.78**).3.item location statistics have direct correspondence to item content, items describing low self-esteem have small Negative location parameter; items describing middle self-esteem have a location parameter which approached to 0;items describing high self-esteem have great location parameter.4.the formal scale have reserved some middle trait items, which describe the state of middle level and have location parameter approached to 0;5.though the results of estimating the RSES, we found that traditional method indeed deleted the items which located in middle location, and reserved the two ends of traits.