Comparison two regressions

#1
Hello everybody, I have a problem... even it is quite diffucult to explain it but I will try:

- I going to carry out a regression analysis of a set of potentially explanatory variables (banks applications - so to 1 application belong approx. 20 different variables)
- I have history of the applications and variables from 2014 and 2018
- I know that at the end of 2015 there were some minor/major modifications in the filling of the variables or at least in some of them (for example, a variable "Expenditures" started from the end of 2015 include higher values in comparison with "Expenditures" to the end of 2015 because of some setting in policy rules or something...but this is not the rule for every one observation, only for some of them and I am not able to find out for which ones)
- The aim is how to find out/affirm that this "change" in the end of 2015 was not significant and I can use the whole dataset from 2014 to 2018 and not only a dataset from the end of 2015 to 2018?

Thank you very much for your ideas and help
 

noetsi

Fortran must die
#2
In time series when you think there is a basic change in the relationship between two variables at a certain point in time you can test this with a chow test. Essentially that shows if there is such a change or not. I am not sure that works for what you want, but you might look it up.
 
#3
In time series when you think there is a basic change in the relationship between two variables at a certain point in time you can test this with a chow test. Essentially that shows if there is such a change or not. I am not sure that works for what you want, but you might look it up.
Thank you for reply but this is not actually a time series yet cross sectional dat. One observation = one application.
 

hlsmith

Not a robit
#4
@tradermartin - I think you need to do a better job describing the context then. Since I interpreted it as time series or a series of cross-sectional data. You can also provide a data sample of real or made-up data to help explain what you are working with data-wise.