Hi all,
Hoping someone may be able to help with the following:
I'm running an OLS regression on a matched sample.
I have an overarching variable to capture my 'treatment' variable which is binary. I'm looking at certain challenges affecting R&D.
That overarching variable is made up of 4 variables, each of which is also binary. There is a correlation among these 4 variables.
When they are included in the model together, the results differ from when they are only accounted for individually. I assume this is down to them having a joint effect on the outcome.
Is there any benefit to running a separate OLS for each challenge? If we run them all together in one model, we can get the effect of one challenge controlling for the others, accounting for the fact they may be jointly influencing the outcome. At the same time if there isn't great variability and there is correlation, would that influence my standard errors and make it harder to see the effect?
I hope this makes sense.
Thank you.
Hoping someone may be able to help with the following:
I'm running an OLS regression on a matched sample.
I have an overarching variable to capture my 'treatment' variable which is binary. I'm looking at certain challenges affecting R&D.
That overarching variable is made up of 4 variables, each of which is also binary. There is a correlation among these 4 variables.
When they are included in the model together, the results differ from when they are only accounted for individually. I assume this is down to them having a joint effect on the outcome.
Is there any benefit to running a separate OLS for each challenge? If we run them all together in one model, we can get the effect of one challenge controlling for the others, accounting for the fact they may be jointly influencing the outcome. At the same time if there isn't great variability and there is correlation, would that influence my standard errors and make it harder to see the effect?
I hope this makes sense.
Thank you.
Comment