# What is the difference between multiple regression & mutivariate regression?

I have data on GDP growth as a dependent variable and growth in main production sectors of Pakistan such as mining, electricity, communication, manufacturing and electricity. I am supposed to run a regression on it. Is there any difference between multiple regression and multivariate regression? If so, than what is it? Which is suitable for my data?

-
The answer to your question appears when you hover your cursor over the multivariate-regression tag. Also see the tag wiki. –  Nick Stauner Apr 17 at 18:52
Furthermore, suitability in this case is less a matter of your data than a matter of the question you want to ask of it. –  Nick Stauner Apr 17 at 19:03

"Multiple regression" refers to situations in which you have more than one predictor / explanatory variable ($X$).

"Multivariate regression" refers to situations in which you have more than one response / outcome / dependent variable ($Y$).

It is also possible to have both multiple predictors and multiple responses, in which case you could call it a "multivariate multiple regression". But since people rarely have only one predictor, I don't think people are worried about making the multiple predictor part distinct. This raises the question of why we worry about "multiple" vs. "simple" (only one predictor) regression in the typical case when you have only one response. I think that it is mostly for historical and pedagogical (teaching) reasons: simple regression was worked out first, and is taught first to help students get the main ideas before going further.

In your case, I gather you have only one response variable (Pakistan's GDP growth), and several predictor variables (growth in mining, electricity, communication, manufacturing and electricity), so your regression model will be a regular old multiple regression.

-
I'm surprised by the definition of "multivariate". I've often heard people employ "multivariate" when they mean "more than one predictor", which apparently isn't correct. –  landroni Apr 17 at 15:08
"situations in which you have more than one response / outcome / dependent variable (Y)." - these are called multinomial. –  Aksakal Apr 17 at 15:16
@landroni, you're right that people often use the term that way, but it isn't correct. (Although words end up meaning what people use them to mean via linguistic drift, so maybe some day it will be.) –  gung Apr 17 at 15:51
@Aksakal, I have never heard the word multinomial used that way in English. When you are predicting 1 of several classes given some information about predictor variables, you can use multinomial logistic regression though, so that may be what you are thinking of. But we still say that you have only 1 response variable, just one w/ multiple levels. –  gung Apr 17 at 15:56