0

Have a look at the plot below (let's assume the vertical axis is called 'y' and the horizontal axis is called 'x'). This is how my two variables are related. I was wondering whether it would be possible to apply a transformation on one of my variables in order to get a more linear relationship? I tried taking the log, but this still gives poor results. The (shortened) dataset looks like this

x    y
1    21.905933
2    12.341214
5    6.601826
10   4.687770
25   3.541561
50   3.224406
100  3.082521
250  2.999060

The goal is to create a linear model with y as dependent variable and x as independent variable.

enter image description here

user3387899
  • 351
  • 4
  • 11
  • 2
    What does this curve look like to you? Also, is this an assignment? If so, please add the self study tag. – jlimahaverford Sep 11 '15 at 14:20
  • log difference? – BCLC Sep 11 '15 at 15:31
  • plot $\log(y-2.95)$ vs $\log(x)$ (or thereabouts) ... but note that this shift-value is effectively an estimated parameter – Glen_b Sep 11 '15 at 16:45
  • 1
    There are some good answers to similar questions [here](http://stats.stackexchange.com/questions/35711/box-cox-like-transformation-for-independent-variables) and [here](http://stats.stackexchange.com/questions/24227/how-to-choose-the-best-transformation-to-achieve-linearity). – Tony Ladson Sep 13 '15 at 00:39

0 Answers0