Problem of the individual change in pretestª²posttest design

    DU Xiaoª²Han£¬ XU Yongª²Yong, CHEN Changª²Sheng

    Department of Health Statistics, School of Preventive Medicine, Fourth Military Medical University, Xi¬ðan 710033, China

    ¡¾Abstract¡¿ AIM: To study the individual change over time in pretestª²posttest design. METHODS: Using the banting of obese children as an example, there were 2 groups: one is the treatment group with the obese children receiving drug to control weight, the other is the control group with the children of standard weight. At the age of 8 and 10, the weight was measured. The hypothesized mean of each group children weight was Y1(t1)=31.0, Y0(t1)=23.0£¬ Y1(t2)=36.0£¬ Y0(t2)=28.0. In the absence of treatment, the hypothesized mean Y1*(t2) was 41.0. RESULTS: Using simple gain scores to estimate the treatment effect, ¦Á=36.0-31.0-(28.0-23.0)=0 (kg), this indicated that after experiment the 2 groups were not different in children weight, and the treatment effect didn¬ðt exist. But the true treatment effect was Y1(t2)-Y1*(t2)=36.0-41.0=-5.0(kg), which showed that the treatment effect existed and the children weight was different after treatment. CONCLUSION: The 4 kinds of individual changing trends in pretestª²posttest design show the problem of the individual change can¬ðt be ignored when we analyze the data of pretestª²posttest design. ......