Hello Usama Waqar so simple

Look at SPSS outputs. You have a table called Classification Table.

Based on this table and the following relation and the attached image, you can calculate RR.

RR = ((a/a+b) / (c/c+d))

Of course, it is clear that you can calculate RR with a lot of software.

Usama Waqar, I agree with Abolfazl Ghoodjani about formula of relative risk (RR). But make sure that you have enough information to calculate RR instead of Odds Ratio (OR). For instance, formula of RR requires the use of all “people at risk” as the denominator. In addition, "Multiple logistic regression, a frequently used multivariate technique, calculates adjusted ORs and not RRs". Look at the Section WHEN SHOULD ONE USE RISK RATIO AND ODDS RATIO? of the following article.

Article Common pitfalls in statistical analysis: Odds versus risk

The UCLA website has pages showing how to use GzLMs to estimate models yielding RRs rather than ORs:

• https://stats.idre.ucla.edu/stata/faq/how-can-i-estimate-relative-risk-using-glm-for-common-outcomes-in-cohort-studies/
• https://stats.idre.ucla.edu/sas/faq/how-can-i-estimate-relative-risk-in-sas-using-proc-genmod-for-common-outcomes-in-cohort-studies/

Unfortunately, neither of them has shows SPSS code. But the models estimated using the -glm- command on that first can be estimated using the GENLIN command in SPSS, I should think. So the Stata & SAS examples might give you some good ideas about how to proceed with SPSS. HTH.

• https://www.ibm.com/docs/en/spss-statistics/27.0.0?topic=reference-genlin

PS- Here's another resource you might find useful:

• https://www.cmaj.ca/content/cmaj/suppl/2011/12/12/cmaj.101715.DC1/risk-knol-2-at.pdf

It is an appendix to this CMAJ article:

Knol, M. J., Le Cessie, S., Algra, A., Vandenbroucke, J. P., & Groenwold, R. H. (2012). Overestimation of risk ratios by odds ratios in trials and cohort studies: alternatives to logistic regression. Cmaj, 184(8), 895-899.

https://www.cmaj.ca/content/184/8/895.short

Just a question on terminology. When people say they want to run a logistic regression, is the only real aspect of this that is required is the logit link. So the question is basically how to tweak the rest of the procedure so that the RRs come out of the model estimates rather than a function of the ORs? Am I interpreting this question correctly, so it is not how to do a GLM with and get these, but a GLM with the logit link.

Have a look at this question:

https://www.researchgate.net/post/What_is_the_difference_between_the_log-binomial_regression_model_and_the_logistic_regression_model

My suggestion is to use negative binomial regression instead of logistic regression

It depends on the nature of your outcome variable. If the outcome variable is continuous variable, it is better to analyse using poison or negative binomial regression model. But if your outcome variable is categorical, you better use Robust poison or Log binomial regression model. Generally, binary logistic regression model is not recommended for Cohort/experimental studies.