I have applied binary logistic regression. My independent variable is bullying victimization (0/1), dependent variable is self-cutting (0/1), covariates are age, gender, school performances, depression, family relationship and so on. In model I in adjusted only bullying victimization and self-cutting, i.e. crude odds ratio 2.27 (1.52-3.39). In model II I adjusted age, gender and bullying victimization 2.89 (1.87-4.47), in model III age, gender, bullying victimization and school performance 2.73 (1.74-4.28), and like that.

However, when I adjusted depression, age, gender and bullying victimization it shows not statistically significant association i.e. 1.40 (0.85-2.31). And because of the depression, when I adjusted all the covariates together in my final model, the association between my independent variable (BV) and self-cutting is not statistically significant.

So how to interpret this particular odds ratio and CI 1.40 (0.85-2.31) which was after adjusting depression in the model? I cannot exclude or remove depression as it is my main covariate that is why I didn't build models with stepwise forward and backward, instead I chose subsets.

When the results is statistically significant it is easier to interpret but how to interpret the results when the association is not significant?

Model I 2.27 (1.52-3.39)    <0.001
Model II    2.89 (1.87-4.47)    <0.001
Model III   2.73 (1.74-4.28)    <0.001
Model IV    2.48 (1.58-3.91)     <0.001
Model V     1.92 (1.21-3.03)     0.005
Model VI    2.36 (1.65-4.29)    <0.001
Model VII   3.69 (2.29-5.95)    <0.001
Model VIII  2.10 (1.27-3.34)    0.003
Model IX    1.40 (0.85-2.31)    0.186
Model X     1.62 (0.89-2.93)    0.112

Model I: Crude odds ratio for the association of self-cutting with bullying victimization Model II: Adjusted for age and gender Model III: Adjusted for age, gender, and school performance (mathematics, science, native language (Finnish) and general subjects) Model IV: Adjusted for age, gender, and fear of going to school Model V: Adjusted for age, gender, and social network-related factors (number of friends, feeling lonely) Model VI: Adjusted for age, gender, and family-related factors (parents’ marital status, relationship with parents, relationship with siblings) Model VII: Adjusted for age, gender, and lifestyle risk factors (use of drugs, use of cannabis, smoking, AUDIT-C) Model VIII: Adjusted for age, gender, and A-DES score Model IX: Adjusted for age, gender, and BDI score Model X: Adjusted for all the variables


1 Answer 1


The meaning of the odds ratio does not change based on whether it is significant or not. It is still an estimate of the ratio of the odds, after adjusting for the other variables in the model.

  • $\begingroup$ Well if the p value and CI is significant then we can interpret that the association is statistically significant, obviously odds ratio might changes slightly. But what I do not understand is that the association between DV and IVs were statistically significant in the model until depression was added. When I added the depression the association became insignificant (p value less than 0.05). So what does this mean? Does it mean that depression has stronger association with dependent variable rather than the independent variable? $\endgroup$ Commented Jun 15, 2023 at 10:55
  • 2
    $\begingroup$ The strength of the association is measured by the OR (or, more generally, its effect size). The significance is measured by p value. Both can change when you add another variable, unless those two variables are orthogonal. Depression is related to all of your other IVs, but especially to victimization. So, some of the variance in self-cutting that was accounted for by victimization is now accounted for by depression. This happens a LOT in all sorts of observational studies, and particularly in psychology and related fields, where IVs are almost always related. $\endgroup$
    – Peter Flom
    Commented Jun 15, 2023 at 11:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.