emmeans: Estimated Marginal Means, aka Least-Squares Means

name: Bug report about: Create a report to help us improve title: '' labels: bug assignees: ''

Please replace the descriptions below with your content that relates to that topic.

Some model classes have their emmeans support in another package -- e.g., the one that defines the model class itself. If so, is the bug really in that package, rather than in emmeans?

Often there are useful annotation that you'll miss if you post-process your code. Seeing the actual results may even explain what's happening!

# --- Actual results (YES):
emmeans(mod, ~ Sex)
##  Sex emmean    SE  df asymp.LCL asymp.UCL
##  F   -1.432 0.582 Inf    -2.572    -0.291
##  M    0.392 0.492 Inf    -0.573     1.356
## 
## Results are averaged over the levels of: Treatment 
## Results are given on the logit (not the response) scale. 
## Confidence level used: 0.95

# --- Filtered results (NO):
emmeans(mod, ~ Sex) %>% as_tibble()
## # A tibble: 2 × 6
##   Sex   emmean    SE    df asymp.LCL asymp.UCL
##   <fct>  <dbl> <dbl> <dbl>     <dbl>     <dbl>
## 1 F     -1.43  0.582   Inf    -2.57     -0.291
## 2 M      0.392 0.492   Inf    -0.573     1.36

A clear and concise description of what the bug is.

Show me code and output that reproduces the bug. Please use a small data set (built-in if possible) and simple variable names. And please, do not create different objects having the same name; to compare two or three different models or methods, give those objects different names so we can talk about them.

Show the actual output from whatever function is in question -- do not pipe it into some post-processing stuff. The actual output often shows some annotations that you don't see if you are trying to "tidy" it. Nothing would make me happier than to avoid all tidyverse code.

A clear and concise description of what you expected to happen.

Add any other context about the problem here.

I really do expect you to look at the documentation and vignettes before sending bug reports. Make sure that you have used things as they are documented.
I really do not want to see your whole workflow. Just show me the code for fitting the model in question (yes, all of the code for that including what libraries are needed), and complete output for where the bug occurs. Leave out things like plots and code/output from other packages unless it is relevant to reproducing the bug.
Show output as pre-formatted text, not a graphic screen shot.
Again, do not ever re-use object names. If you are comparing results for two or more models or datasets, assign them different names.
More than one or two pipes is usually too many. I'd rather see the individual steps and the actual, unadorned results thereof.
Did you know that there is an index of vignette topics? That can be helpful for clarifying certain issues. See https://cran.r-project.org/web/packages/emmeans/vignettes/vignette-topics.html
Please examine the direct output from what you have tried. Often there are a few lines of annotation below the output. If you wrap or pipe your results with some kind of post-processing, you suppress those annotations. I consider that willful ignorance, and will not help you.