I tried 3 times until now.
when odds ratio is not possible due to 0 outcome risk estimate can be calculated for the control group ( SPSS, descriptive statistics, crosstabs, risk). So how I should understand the risk value form the the tables?
Can you provide your contingency table data. Is it possible you have not sample enough people?
A possible fix is to add a nominal amount to all of the cells, typically 0.5 (referred to as acontinuity correction). You would just want to disclose the use of this approach when reporting the results.
Sorry. as I see I am not successful with sending my first reply.
Here is it.
Yeah I do not have enough people. But the problem with the data is the outcome is 0 for some outcomes under the same risk factor. So SPSS did not calculate odds ratio for these outcomes and jumped to Risk Estimate which is even harder to understand for me.
Now do you mean I must add 0.5 to all cells even to those which caused 0 outcome on the crosstabs?
Thank you very much! I understand what to do. I will add this value just to calculate the risk and report it in the statistical analysis section only for the odds ratio calculation. I did all the previous analysis before fixing the values used to calculate the odds ratio. As a statistical approach, is that fine?
Hi, Although I added 0.5 to all the cells the result or risk estimate did not changed. I got the same results: 0 still exist on the crosstabs and odds ratio was not calculated for the risk. Only risk estimate for 0,5= no outcome was calculated...
There is no attachment button here to send the analysis as I see...
I calculated correlation seperately between 6 risk factors and 5 outcomes. Then I needed to add odds ratio.
Yes it is a 2×2 table.
I can see the risk estimate include odds ratio. but it is different from it....
They are not the same. If the exposure of interest is rare, then the odds can approximate the risk. Risk is usually used if you have temporality in data (e.g., collect exposure then document subsequent outcome) and odds if data collect retrospectively or via cross-section without knowledge of temporality. Odds and odds ratios get used more frequently since they can be easily generated from logistic regression.