Conditional Logistic Regression

 

The Basic

 

Conditional logistic regression is useful in investigating the relationship between an outcome and a set of prognostic factors in a matched case-control studies.

 

Matched case-control study designs are commonly implemented in the field of public health. While matching is intended to eliminate confounding, the main potential benefit of matching in case-control studies is a gain in efficiency.

 

When there is one case one control in a matched set, the matching is 1:1. When there is one case and a varying number of controls in a matched set, the matching is 1:n. For 1:n set, conditional logistic regression is appropriate to model the outcome (case or control) with independent variables.

 

Select matched set:

 

  Usually a variable is designate to indicate the matched set. Select this variable in “Select matched set” box.

 

 

Below is the sample input window

SNAGHTML11809da8

 

 

Below is the sample output and explanation of the above model:

 

Conditional logistic model

 

(1)     First, output model formula

 

Call:

coxph(formula = Surv(rep(1, 248L), CASE) ~ SPONTANEOUS + INDUCED +

    strata(STRATUM), data = WD, method = "exact")

 

  n= 248, number of events= 83

 

(2)     Next, output model regression coefficient and p value

 

              coef exp(coef) se(coef)     z Pr(>|z|)   

SPONTANEOUS 1.9859    7.2854   0.3524 5.635 1.75e-08 ***

INDUCED     1.4090    4.0919   0.3607 3.906 9.38e-05 ***

---

Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

 

(3)     Next, output odds ratio and 95% confidence interval for each independent variable

 

            exp(coef) exp(-coef) lower .95 upper .95

SPONTANEOUS     7.285     0.1373     3.651    14.536

INDUCED         4.092     0.2444     2.018     8.298

 

(4)     Next, output model diagnostic parameter

 

Rsquare= 0.193   (max possible= 0.519 )

Likelihood ratio test= 53.15  on 2 df,   p=2.869e-12

Wald test            = 31.84  on 2 df,   p=1.221e-07

Score (logrank) test = 48.44  on 2 df,   p=3.032e-11