Graph to show non linear relationship

[Faisal Abdullah]

Hello Everyone,

In the xtreg regression analysis,

Y= Leverage X1= Size X2 = Size^2

I find that both size and also size square are statically significant. Size has negative coefficient but size square show positive coefficient. the coefficients values are -1.219097 for size and 2.15018 for the variable size square. Is this okay?


is the multicollinearity an issue between a variable and the square of the same variable ?


How I can show this non liner relationship in graph? should I plot size or size square?

I tried many times to plot the relationship correctly but I could not.

If I find in the regression the size square is non linearly related to leverage, does this mean that this relationship will always revealed in a graph?

Many thanks and regards for you support

Faisal

[Rich Goldstein]

you don't say whether you have "centered" the size variable - doing that will help take care of any collinearity issue; note that, based on the minimal info you provide, there is not an issue here anyway

you don't want to plot just size or just size-squared; assuming you have used factor variable notation as in

xtreg leverage c.size##c.size

probably your easiest way to get a plot is to use the user-written mcp command; use search to find and install:

search mcp

[Nick Cox]

regplot from SJ will do this assuming that your code resembles Rich's.

search regplot

and download from the most recent source.

This works:

 
webuse grunfeld
xtreg mvalue c.invest##c.invest 
regplot

although don't ask me what sense that makes.

Note that your relationship is a quadratic from a linear model.