# Controlling for national wages

#### noetsi

##### Fortran must die
Not a statistics problem in the classical sense but I could use help on this. Our wages (I work for VR) for our customers are below the national average. I think this is largely or entirely due to the fact that wages in Florida generally are lower than other states. Which drives our customer's wages down.

But I can't think of how to control for this.

#### spunky

##### Doesn't actually exist
Well, what are you trying to do with this wage control situation of yours? In my experience, most of the time you hear "we controlled for wages" or "we controlled for gender" usually means you threw it in as a covariate in some sort of regression model and you're "controlling for" it because you're not actually interested in saying anything about it, but rather make sure the explained variance (in an OLS context) is due to the variables you'd like to explain and not to control for.

#### hlsmith

##### Not a robit
Yeah, it comes down to what you are trying to generate at the end. Are you trying to compare your outputs to another dataset or values. Definitely an economics question. It feels like a correction could be added at the end of the modeling, or can you extrapolate your findings using other data in your final model such as heights predict weights, I had not 7 foot tall people in my set but this is what their predicted weights would be, given a linear relationship exists.

But is comes back to what you are trying to do.

#### noetsi

##### Fortran must die
In this case I am not running a regression model. I am trying to adjust our number to address differences between our state and the nation.

I have our ratio of customer wages compared to other VR customer wages and I am trying to adjust it for the difference for wages generally in our state (which are low) to the national wages. My theory is our customers earn less than other VR agencies because wages are lower generally here. But I am not sure how I can adjust the number to address these differences.

#### hlsmith

##### Not a robit
Do you have the values? In medicine disease and death rates are traditionally adjusted using standardization approaches, (e.g., deaths per 1,000 or say put things onto a standard normal distribution.

#### noetsi

##### Fortran must die
I can get the values but I am not trying to standardize in that regard. It is already a ratio (our level/their level). I just need to adjust for the differences in labor markets between the US and Florida. That is a form of standardization I guess.

This is what I am talking about. Say our wages are 15 dollars an hour and the average for the VR organizations was 20. So the ratio is 3/4 - bad for us. But Florida wages in general are 10 hour and national wages are 15. How would I adjust our numbers to show that Florida wages were lower generally than the US.