The package EffectStars2 provides a more up-to-date implementation of effect stars!
The function computes and visualizes cumulative logit models. The computation is done with help of
the package VGAM. The visualization is based on the function stars from the package graphics.
star.cumulative(formula, data, global = NULL, test.rel = TRUE, test.glob = FALSE,
partial = FALSE, globcircle = FALSE, maxit = 100, scale = TRUE,
nlines = NULL, select = NULL, dist.x = 1, dist.y = 1, dist.cov = 1,
dist.cat = 1, xpd = TRUE, main = "", col.fill = "gray90",
col.circle = "black", lwd.circle = 1, lty.circle = "longdash",
col.global = "black", lwd.global = 1, lty.global = "dotdash", cex.labels = 1,
cex.cat = 0.8, xlim = NULL, ylim = NULL)An object of class “formula”. Formula for the cumulative logit model to be fitted and visualized.
An object of class “data.frame” containing the covariates used in formula.
Numeric vector to choose a subset of predictors to be included with global coefficients. Default is to include all coefficients category-specific. Numbers refer to total amount of predictors, including intercept and dummy variables.
Provides a Likelihood-Ratio-Test to test the relevance of the explanatory covariates.
The corresponding p-values will be printed as p-rel. test.rel=FALSE might
save a lot of time. See also Details.
Provides a Likelihood-Ratio-Test to test if a covariate has to be included as a category-specific covariate (in contrast to being global). The corresponding p-values will be printed as p-global. test.glob=FALSE and globcircle=FALSE might save a lot of time. See also Details.
If partial=TRUE, partial proportional odds models with only one category-specific covariate are fitted.
The resulting effects of the (sub)models are plotted. For further information see Details.
If TRUE, additional circles that represent the global effects of the covariates are plotted. test.glob=FALSE and globcircle=FALSE might save a lot of time.
Maximal number of iterations to fit the cumulative logit model. See also
vglm.control.
If TRUE, the stars are scaled to equal maximal ray length.
If specified, nlines gives the number of lines in which the effect stars are plotted.
Numeric vector to choose only a subset of the stars to be plotted. Default is to plot all stars. Numbers refer to total amount of predictors, including intercept and dummy variables.
Optional factor to increase/decrease distances between the centers of the stars on the x-axis. Values greater than 1 increase, values smaller than 1 decrease the distances.
Optional factor to increase/decrease distances between the centers of the stars on the y-axis. Values greater than 1 increase, values smaller than 1 decrease the distances.
Optional factor to increase/decrease distances between the stars and the covariates labels above the stars. Values greater than 1 increase, values smaller than 1 decrease the distances.
Optional factor to increase/decrease distances between the stars and the category labels around the stars. Values greater than 1 increase, values smaller than 1 decrease the distances.
If FALSE, all plotting is clipped to the plot region, if TRUE, all plotting is clipped to the figure region,
and if NA, all plotting is clipped to the device region. See also par.
An overall title for the plot. See also plot.
Color of background of the circle. See also col in par.
Color of margin of the circle. See also col in par.
Line width of the circle. See also lwd in par.
Line type of the circle. See also lty in par.
Color of margin of the global effects circle. See also col in par. Ignored, if globcircle = FALSE.
Line width of the global effects circle. See also lwd in par. Ignored, if globcircle = FALSE.
Line type of the global effects circle. See also lty in par. Ignored, if globcircle = FALSE.
Size of labels for covariates placed above the corresponding star. See also cex
in par.
Size of labels for categories placed around the corresponding star. See also cex
in par.
Optional specification of the x coordinates ranges. See also xlim in plot.window
Optional specification of the y coordinates ranges. See also ylim in plot.window
P-values are only available if the corresponding option is set TRUE.
Odds or exponential coefficients of the cumulative logit model
Coefficients of the cumulative logit model
Standard errors of the coefficients
P-values of Likelihood-Ratio-Tests for the relevance of the explanatory covariates
P-values of Likelihood-Ratio-Tests wether the covariates need to be included category-specific
xlim values that were automatically produced. May be helpfull if you want to specify
your own xlim
ylim values that were automatically produced. May be helpfull if you want to specify
your own ylim
The underlying models are fitted with the function vglm from the package VGAM. The family argument
for vglm is cumulative(parallel=FALSE).
The stars show the exponentials of the estimated coefficients. In cumulative logit models the exponential coefficients can
be interpreted as odds. More precisely, the exponential \(e^{\gamma_{rj}}, r=1,\ldots,k-1\) represents the multiplicative effect of the covariate j on the cumulative odds \(\frac{P(Y\leq r|x)}{P(Y>r|x)}\) if \(x_j\) increases by one unit.
In addition to the stars, we plot a cirlce that refers to the case where the coefficients of the corresponding star are zero. Therefore, the radii of these circles are always \(exp(0)=1\). If scale=TRUE, the stars are scaled so that they all have the same maximal ray length. In this case, the actual appearances of the circles differ, but they still refer to the no-effects case where all the coefficients are zero. Now the circles can be used to compare different stars based on their respective circles radii. The p-values beneath the covariate labels, which are given out if test.rel=TRUE, correspond to the distance between the circle and the star as a whole. They refer to a likelihood ratio test if all the coefficients from one covariate are zero (i.e. the variable is left out completely) and thus would lie exactly upon the cirlce.
The form of the circles can be modified by col.circle, lwd.circle and lty.circle.
By setting globcircle=TRUE, an addictional circle can be drawn. The radii now correspond to a model, where the respective covariate is not included category-specific but globally. Therefore, the distance between this circle and the star as a whole corresponds to the p-value p-global that is given if test.glob=TRUE.
Please note:
Regular fitting of cumulative logit models may fail because of the restrictions in the parameter space that have to be
considered. If partial=TRUE, (sub)models with only one category-specific covariate, so-called
partial proportional odds models, are fitted. Then at least estimates for every coefficient should be available. If partial=TRUE, the resulting effects of these (sub)models are plotted.
It should be noted that in this case no coherent model is visualized. Also the p-values refer to the various submodels.
For partial=TRUE, the p-values p-rel and p-global refer to tests of the corresponding partial proportial odds models against the proportional odds model.
It is strongly recommended to standardize metric covariates, display of effect stars can benefit greatly as in general differences between the coefficients are increased.
Tutz, G. and Schauberger, G. (2012): Visualization of Categorical Response Models - from Data Glyphs to Parameter Glyphs, Journal of Computational and Graphical Statistics 22(1), 156-177.
Gerhard Tutz (2012): Regression for Categorical Data, Cambridge University Press
# NOT RUN {
data(insolvency)
star.cumulative(Insolvency ~ Sector + Employees, insolvency, select = 2:4)
# }
Run the code above in your browser using DataLab