0
$\begingroup$

I am doing an analysis using logistic regression and my odds ratio is going pretty high. When I use a combination of the three IVs, with three other control variables, odds ratio for one of the IVs is 62, which I guess is pretty high.

        Variables in the Equation
              B     S.E.    Wald    df  Sig.    Exp(B)
FirmSize    .001    .012    .005    1   .942    1.001
FirmAge     -.048   .059    .680    1   .410    .953
FirmI       .035    .127    .076    1   .783    1.035
TR          -1.190  .861    1.910   1   .167    .304
ITR         4.132   1.417   8.498   1   .004    62.301
EO          2.356   .911    6.697   1   .010    10.553
EU          -.270   .577    .219    1   .640    .764
Constant    -17.535 5.298   10.955  1   .001    .000

a. Variable(s) entered on step 1: Tangibleresource, Intangibleresource, EO, EU.

Independently if I test each variable, the odds ratio is even bigger, considering the impact of only one variable being taken into account.

My friend suggested that scale suggested that the scale used would be inappropriate and asked me to convert the units. But my IVs, all three of them, are recorded on likert scale (1-5)so it's not like there is inconsistency in measurement approach.

What should I do?

$\endgroup$
  • $\begingroup$ How many cases do you have in the lowest-frequency outcome category? $\endgroup$ – EdM Apr 1 '17 at 16:27
  • $\begingroup$ You could be overfitting. Maybe you should remove some of the variables. $\endgroup$ – Michael Chernick Apr 1 '17 at 16:29
  • $\begingroup$ Hi EdM, the outcomes are two (0 and 1). 0 has 70, 1 has 57. Total 127 observations $\endgroup$ – Rohit Apr 1 '17 at 17:22
  • $\begingroup$ Hi Michael, even if I tried just one variable, without anything else, odds ratio goes pretty high. Is it because of DoF that you suggested that? $\endgroup$ – Rohit Apr 1 '17 at 17:23
  • $\begingroup$ Check whether your ITR variable has much variation. Maybe nearly all responses are "5" and nearly all the non-"5"s have a different outcome from most "5"s. $\endgroup$ – rolando2 Apr 1 '17 at 20:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.