Question

我知道如何使用Rattle with R通过在模型和验证案例中分离数据集来应用logit模型。我是否可以在SAS中获得任何明确的指导/信息来源。有可能使用Proc Score和Proc logistic ......以下某种方式方式...但我很困惑

＆lt;＆lt;＆lt;＆lt;＆lt;＆gt;＆gt;＆gt;代码块＆lt;＆lt;＆lt;＆lt;＆gt;＆gt;＆gt;＆gt;＆gt;

proc logistic data=logistic descending;
 model credit = &varlist;
 output out=out1 predprobs=(i);
 score data=new out=out2;
 run;

proc print data=out1(obs=n);
 run;
 proc print data=out2;
 run;

proc logistic inmodel = model; 得分数据= new out = out2; 运行;

proc print data = out2; 运行;

Answer 1

proc logistic data=train outest=est;
model y=x;
run;

proc score data=test score=est type=parms out=test2;
var x;    
run;

data test2;
set test2;
prob=exp(y)/(1+exp(x));
run;

proc means;
run;

Answer 2

/* develop model */
proc logistic data= train_data desc;
    model response_var = <var list>;
    /* output scored model into dataset */
    output out= <train_data2> predicted= <name of predicted var - eg: p_hat>;
    /* apply score to validation dataset */
    score data= test out= test2;
run;

通过PROC SCORE还有另一种方法可以做到这一点，但我无法弄明白。

然后，您需要在test2数据集上运行诊断/模型评估。
例如：

proc rank data= test2 groups=10 out = test3 descending ties = high;
    var P_1;
    ranks pred_v1;
run;


proc sql;
    select pred_v1, sum(response_var) as resp,  count(*) as count
    from test3
    group by pred_v1
    order by pred_v1 asc;
quit;

显然，您可以使用多种诊断方法;但是，显示此示例是为了完整性。

通过在SAS中分离模型和验证案例中的数据集来应用logit模型

2 个答案: