Hello every one,
I run binary logistic regression in SPSS but i did not have results because of complete separation. How can i solve this proplem?
Thanks in advance.
Complete separation in logistic regression occurs if an independent variable perfectly predicts the outcome variable. This means that there is a linear combination of the predictor that can separate the cases where the outcome is 1 completely from those where the outcome is 0. When this happens, the maximum likelihood estimates for the coefficients in logistic regression become infinite, causing standard logistic regression methods to fail.
There are ways to handle complete separation in logistic regression in SPSS, although the capabilities may be more limited compared to specialized statistical software such as R. The simplest method you can do in SPSS is to recode the respective predictor variable(s) in such a way that you combine categories that can lead to separation. All other methods include installing extensions such as “R” or “Python” in SPSS to call R packages that can handle complete separation. For example, you can use R scripts for Firth's penalized likelihood regression ('brglm') or Bayesian approaches ('rstanarm') to solve or mitigate the problem of infinite coefficients caused by full separation in SPSS.
If you use SPSS 29, the R integretion plug-ins should already be part of the SPSS installation. Look here:
https://www.ibm.com/docs/en/spss-statistics/saas?topic=r-
In older versions, such as SPSS 25, you have to install it manually, following a certain procedure, as described here:
https://www.ibm.com/support/pages/r-installation-ibm-spss-statistics-25-client-and-server
Hope this helps.
Regards
If you need to install it manually, you can do so fairly easily via the Extension Hub: Extensions > Extension Hub. Then search on "Firth", as in the attached image. HTH.
Bruce Weaver
Thanks alot
When i did that, the tab contain only installed part not contain explore
Samah Mohammed Refay, I don't understand what you mean. If you mean you are not seeing the extension command, try checking all 3 boxes in the bottom left corner:
- Installed
- Not installed
- Update available
Bruce Weaver, Dr has any source for performing firth regression and interpreting the result of it in SPSS.
Alexander Pabst
How can i obtain Odd ratio from firth regression in SPSS?
If there is no option to report ORs in the output, OR = exp(coefficient); and the limits of the CI for the OR = exp(lower bound) and exp(upper bound).
Bruce Weaver
How can I calculate adjusted odds ratio for result of multivariate firth logistic regression?
Samah Mohammed Refay, my earlier reply explained how to do this. But here is more detail.
Your output will have a table showing coefficients and CIs. Copy that table using the Copy as Excel Worksheet option and paste it into Excel. Then use the EXP() function in Excel to compute the ORs and their CIs.
- Adj OR = Exp(B), where B is the coefficient
- Lower bound of CI for Adj OR = Exp(lower bound of CI for B)
- Upper bound of CI for Adj OR = Exp(upper bound of CI for B)
Alternatively, you could use the OMS command to write the table of coefficients to a new dataset and then do the needed computations in SPSS. But if you are not someone who uses command syntax regularly, it will probably be easier for you to use Excel.
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Enzymology
- Enzymes