interpreting odds ratios in ordinal logistic regression

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

hmz73

New Member
7+ Year Member
Joined
Jul 22, 2016
Messages
1
Reaction score
0
Please allow me to introduce my self, I am student who doing a research and I would like to ask you

Which is the best interpretation to explain (OR = 2.76 [2.180 - 3.492] for livestock such as goats, sheep and pigs correlated with Malaria in certain community through cross sectional study, in these sentences below?

Respondents or household who keep livestock such as goats, sheep and pigs have a 2.8 times greater chance of contracting malaria compared to a respondents or household who do not raise cattle where the confidence interval [CI: 2.180 - 3492])

or

By using multiple logistic regression, we concluded that residents who raised cattle, goats, sheep and pigs increased the risk of getting malaria by 2.8 times compared to participants who did not keep cattle (OR = 2.76 [2.180 - 3.492], after controlling for other covariates in the areas under investigation.

or

This implies that the participants or households who raise cattle such as goats, sheep and pigs, respectively, have the increased risk of contracting malaria by 2.8 times higher than participants who do not keep the livestock/pets, after adjusting other predictor variables.

Annotated Stata Output: Ordered Logistic Regression, like attached

Thank you very much for your prompt reply.

Best regards

Members don't see this ad.
 

Attachments

  • finalmodel.JPG
    finalmodel.JPG
    71.7 KB · Views: 41
It means that the independent relationship (ie. if both those that had these animals and did not have these animals were similar on all adjusted factors) is that those with these animals have 2.8 times or 180 percent the odds of being diagnosed/reporting malaria in this particular sample. There is a 95 percent chance that the odds in the target population falls in the confidence interval.
 
Members don't see this ad :)
Be very careful in how you word this; cross-sectional studies cannot definitively declare a cause-effect relationship because of their observational/epidemiologic nature where all confounding and interacting factors cannot be controlled for (think interventional studies within the lab where subjects can be assigned to the control and others to the variable) and outcomes replicated. New hypotheses can be developed from the observational studies like cohort, case-control, and cross-sectional but hard facts will not be obtained for this reason. Thus, you would be less likely to have peer review send it back if you said something along the lines of:

Subjects exposed to livestock including goats, sheep, and pigs were 2.76 times more likely to present with malaria than those not exposed to livestock (OR=2.76, 95% CI 2.180 - 3.492).

No conclusions were made in the statement above stating that the exposure is what caused the increased risk of malaria. It is simply stating that the livestock exposure related to an increase in the outcome (although may not have truly caused it).
 
  • Like
Reactions: 1 users
It means that the independent relationship (ie. if both those that had these animals and did not have these animals were similar on all adjusted factors) is that those with these animals have 2.8 times or 180 percent the odds of being diagnosed/reporting malaria in this particular sample. There is a 95 percent chance that the odds in the target population falls in the confidence interval.
I know this is an old post, but I'm bumping it in case someone else doing research comes across it. The bold is an incorrect, common misinterpretation of a confidence interval (even Wikipedia has this one). Any specific CI contains or does not contain the true parameter value (probability 1 or 0, respectively). The confidence level is a statement about the method's long run performance; for some fixed parameter value, the confidence level represents the proportion of all possible intervals of the same CI level, created by repeated sampling, that capture the true parameter value. You can't make a probability statement about any specific interval, though.
 
Top