3
$\begingroup$

I am using the following equation form for presenting a linear model and also for explaining how well the inputs are affecting the response variable $ua$.

$$ua = (80) + (-13 \cdot ma) + (-0.4 \cdot ka) + (-1 \cdot lb) + (8 \cdot tb) + (-9 \cdot sa)$$

However, I am not happy with this representation on my pdf file (from LaTeX). Is there any better way we can present this using one single figure? This figure should provide all the above details such as coefficients, directions (+/-), along with the adjusted $R^2$ value of the model. Please suggest to me if there is anything for this purpose?

Improve this question
$\endgroup$
1
  • 4
    $\begingroup$ If the variables aren't all on the same scale, then the $\beta$'s will be difficult to represent all on one plot. If the variables are all standardized, for example, you could make a plot with side-by-side dots for the point estimate and error bars around the bars - the numerical values on the $y$-axis and the variable names on the $x$-axis $\endgroup$
    – Macro
    Commented Aug 1, 2011 at 21:41

1 Answer 1

Reset to default
7
$\begingroup$

You might be interested in the paper "Using Graphs Instead of Tables in Political Science". One section of the paper is devoted to "Plotting Regression Coefficients". Andrew Gelman likes this type of plot (coefplot), too (the R code can be found in his arm package, see here for some examples).

However, I agree with @Macro that these plots are not easy to interpret when the coefficients are not on the same scale.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.