I am rather new to regression analysis, having a completely different background. I am trying to build a model for predictions.

The distribution of my dependent variable, Value (in US$), is right skewed and nonnegative. My explanatory variables are CurrentValue, i.e., the value of the item one year before, the age A of the item, and a categorical variable Type.

Reading a bit, also from other posts here, I came to the conclusion that a GLM with gamma distribution and log link seems a suitable choice. In R:

fit <- glm(Value ~ CurrentValue + Age + Type -1, family=Gamma(link="log"),data=values)  

My question, probably trivial to the most, is the following: How do I back-transform the predicted value into the scale of my dependent variable Value? More simply, given that I call:

predicted <- predict(fit,newvalues) 

how do I obtain values in US$, given the vector predicted? Any help will be highly appreciated.


If you include type="response" in your predict code then the predicted values will be on the response scale.

predict(object, newdata = NULL, type = c("link", "response", "terms"), se.fit = FALSE, dispersion = NULL, terms = NULL, na.action = na.pass, ...)

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.