Extremely large values of Odd ratio (Exp (B)) in logistic regression

[MEGHA TYAGI]

Respected Neeraj sir, 

My study includes predicting the probability of a high or low score (DV) due to built environment variables (IDV1, IDV2, IDV3, IDV4). Here, DV is categorical with '0' indicating low score and '1' indicating 'high score'. All IDVs are metric (ratio scale). Through your videos, I have learned that to conduct a bivariate analysis with DV and 1 IDV (IDV1-Land-use mix), logistic regression will be applied here. But, before that, we need to check the relationship between these two variables. So, I checked the correlation value and also conducted independent t-test. Both the results were showing a good correlation (Pearson correlation= 0.523, p-value=0.000; t= -7.319, p-value=0.000, equal variance assumed). 

Next, bivariate logistic regression was conducted (enter method, not included any control variable). Results are as follows:

a) -2LL was initially 191.522, after 6th iteration, the value becomes 130.821

b) Omnibus test of model coefficients, the p-value is significant (p-value=0.000) (chi-square= 60.701)

c) Cox and Snell R square = 0.344 and Nagelkerke R =0.468 

d) Hosmer and Lemeshow test, p-value =0.65

e) Classification table suggests overall prediction is 81.9 % correct

f) Variables in the equation result are as follows:

Variables in the Equation

                           B          S.E.   Wald      df       Sig.     Exp(B)                    95% C.I.for EXP(B)

                                                                         Lower          Upper

Step 1a Land-use mix 22.958 3.932 34.088 1 .000 9348643578.839 4203275.002 20792628777285.350

       Constant        -17.996 3.040 35.052 1 .000 .000

a Variable(s) entered on step 1: Land-use mix.

g) correlation table is as follows:

 Correlation Matrix

Constant Land-use mix

Step 1 Constant 1.000 -.997

Land-use mix -.997 1.000

This was a problem encountered in other cases as well. Such an extremely large value for the odds ratio is a problem here which I am not able to decipher. 

Kindly let me know if I am doing something wrong here or my process is somewhere not appropriate. 

Thanking you

Megha Tyagi

PhD student

Department of Architecture and Planning

IITR