Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

There doesn't appear to be much motivation for using anything other than a linear regression on this, and the result would be a lot flatter than the trend line given.


A linear regression would be theoretically wonky, since the Y-axis is upper bounded at 100%. So, by the time the graph hits values like 80%, it should start to get "saturated" and flatten out. This fit, with A approximately 100 does a good job of that.

Clearly, the fit is bad near the origin (for small purchasing power), so maybe one can expect a modification at small values of x; I'm not sure if that that will correspond to small values of y.


The line of an ordinary least squares fit would go from approximately 50% on the left to 70% on the right or maybe a little bit less steep than that.


Yep, that, and the regression seems to be forced through the origin, which is stupid. And the dependent variable is on the X-axis. All pretty weird choices.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: