How to backtransform data that has been log transformed in order to report raw values for ease of interpretation? I have run some lme4 analyses on reaction time data in R, with RT being the main outcome variable of interest, which I first log transformed due to non-normality that is typical in RT tasks. However, it makes more sense to talk about the data in terms of milliseconds, rather than the log-transformed values. Obviously I cannot simply run the same model with the raw RT data to get the millisecond values, as I get different results wrt 'significance' of various terms in the model. So, I assume I need to do some sort of back transform of the log transformed values, but I am not sure how to do that. 
Background: I am new to R, lme4, and mixed models in general ;-). I have done some searching but fear that it is my lack of use of correct terminology that has kept me from finding the correct procedure. 
 A: To visualize values in a mixed model in the back transformed scale, you first need to build the model with the original data and apply the data transformation as needed. Then, whenever you need to see the back-transformed values, add the argument 'type=response'.
For example, if your model wanted to predict log-transformed values of variable A based on a categorical variable B and corrected for a random factor, the model would be something like:
model <- lmer(log(A) ~ (B) + (1|random_factor), df)

Then, if you run a post-hoc analysis in emmeans, use:
comparison <- emmeans (model, specs = pairwise ~ B, type = 'response')

or if you want to get the predicted values along the categorical variable and the original values, you can run something like:
df_predicted <- data.frame(A_predicted = predict(model, 
  type="response"), df$A, df$B)

I think this answer you question in very general terms, unfortunately, there are very little details for your specific case, but this should be applicable to many situations, which is good.
