jessie31 发表于 2010-10-1 11:24 
我想建一个模型logistic regression model,预测大学新生的毕业率。变量hsgpa (high school GPA), gender 都显著。
goodness fit 结果并不好,Pearson statistic is 0.1025 and  Deviance is only 0.0351.
想请问下大家,为什么 Pearson statistic,  Deviance 两者差别这么大?谢谢大家, 以下是sas output.
Deviance and Pearson Goodness-of-Fit Statistics
Criterion        Value        DF        Value/DF        Pr > ChiSq
Deviance        414.3363        364        1.1383        0.0351
Pearson        398.5674        364        1.0950        0.1025
Number of unique profiles: 367
Note:The covariance matrix has been multiplied by the heterogeneity factor (Deviance / DF) 1.13829.
Model Fit Statistics
Criterion Intercept Only         Intercept and Covariates
AIC        2515.367        2310.963
SC        2521.045        2327.998
-2 Log L        2513.367        2304.963
Testing Global Null Hypothesis: BETA=0
Test                        Chi-Square        DF        Pr > ChiSq
Likelihood Ratio        208.4041        2
The definitions of two statistics are different. See the links below.
http://en.wikipedia.org/wiki/Pearson's_chi-square_test
http://en.wikipedia.org/wiki/Deviance_(statistics)