I am performing a mixed-effects Poisson regression on count data in R and have estimates for an intercept (my control) and two factors. The estimates of the fixed effects are on the log scale, so I took the exponential of them. I am aware that this works well for the estimates but not for their standard errors, so I calculated confidence intervals on the log scale that I then back-transformed by taking the exponential. As the estimates of the fixed effects are relative to the intercept, I multiplied them by the intercept to get absolute estimates for a bar graph (estimated mean counts). But what do I do to calculate the corresponding confidence intervals?
I believe SD for any log scale is .5.
Assum9ng you squared these values, perhaps .025 is your SD estimate for converting to SE.
Is it how to plot CIs?
You can use the function arrows() with code=3, angle=90. This will draw double-T-like error bars. If you don't want the line ends, you can use the function segments().
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs