2

Assuming, I have constructed a simple linear regression model (Time Series) with a dependent and two independent variables. Measuring the effect concerning the value of the dependent variable is pretty straight forward. Now, I am wondering wether there is a method of estimating the effect of each independent variable individually on the variance of the dependent variable.

Thanks in advance

Steffen Moritz
  • 1,564
  • 2
  • 15
  • 22
shenflow
  • 750
  • 8
  • 20
  • 1
    You could try folding the outcome about its mean or median. – mdewey Oct 25 '17 at 14:26
  • Thank you for the answer! Yet, I dont think I completely understand your suggestion in this context. Would you mind specifying? Thank you so much! – shenflow Oct 25 '17 at 14:56
  • If you fold the variable then extreme values on your old one map to high on the new one and values near the middle map to low values. Then you can just treat it as usual. I am not claiming this would be the only or best way which is why it was only a comment not an answer. – mdewey Oct 25 '17 at 15:02
  • Mhm I think you maybe misinterpretated my question.. I am aiming for some kind of assesment like: "the parameter x has a positive effect on the variance of y, more precisely, the fluctuation of x within the respective period accounts for ?? of the variance of y"... and I dont know how to estimate this effect. – shenflow Oct 25 '17 at 19:09
  • 1
    @shenflow It sounds like you are interested the R-squared value decomposed into components of each independent variable: https://stats.stackexchange.com/questions/60872/how-to-split-r-squared-between-predictor-variables-in-multiple-regression – Michael Webb Oct 25 '17 at 19:24

0 Answers0