Il y a près de 7 ans, j'écrivais un article intitulé La vie après le pétrole, après la lecture d'un livre du même intitulé.

7 ans après, que s'est il passé ? Eh bien depuis 2005, la production de pétrole a stagné ! C'est en tous cas ce qu'affirme un commentaire dans Nature, repris par Sandra Hines. Chiffres à l'appui, on découvre le plateau, la tôle ondulée, les prix qui explosent. Faillite de Petroplus, tensions en Iran, tous les ingrédients sont réunis pour le plus gros choc pétrolier de l'Histoire, tel que prédit par Jean-Luc Wingert.

Alors il est vrai que depuis, pas mal de choses ont changé. Les énergies renouvelables ne sont plus une chimère défendue uniquement par les Verts. Nombreux sont ceux qui ont compris que les bénéfices que l'on peut en tirer sont loin d'être uniquement écologiques. Mais le chemin à parcourir pour être totalement indépendant de l'or noir semble encore bien long.

It should be rather transparent, but I just switched my website from ePerl to Catalyst, as announced before. The application should be a bit more reliable (less error 500 pages, that is after the initial debugging period).

pROC's steady progression goes on with version 1.5. It is avalable for R only. S+ users will need to wait for the upcoming 1.6 release which will introduce power / sample size computations.

This version introduces four new notable features:

• Variance and covariance
• Univariate Log-Concave Density Estimation smoothing
• Improvements to the plotting function
• New return values in coords

Variance and covariance

It is now possible to compute the variance of a ROC curve, and the covariance of two paired ROC curves.

library(pROC)
data(aSAH)
rocobj <- roc(aSAH$outcome, aSAH$s100b)

var(roc1)
var(roc2)
cov(roc1, roc2)

Two methods are available: bootstrap, DeLong¹. The bootstrap is the most versatile method. DeLong is faster but works for full AUC only. For more details, see ?var.roc and ?cov.roc.

Univariate Log-Concave Density Estimation smoothing

Until now, three methods were available to smooth a ROC curve: binormal, density and fitdistr (to fit a distribution with MASS). Now, two new methods are available: logcondens and logcondens.smooth. They are based on Duembgen and Rufibach (2011)². You first need to install the logcondens package:

install.packages("logcondens")

It doesn't need to be loaded.

plot(rocobj)
rs <- smooth(rocobj, method="binormal")
plot(rs, add=TRUE, col="green")
rs2 <- smooth(rocobj, method="density")
plot(rs2, add=TRUE, col="blue")
rs3 <- smooth(rocobj, method="fitdistr", density="lognormal")
plot(rs3, add=TRUE, col="magenta")
rs4 <- smooth(rocobj, method="logcondens")
plot(rs4, add=TRUE, col="brown")
rs5 <- smooth(rocobj, method="logcondens.smooth")
plot(rs5, add=TRUE, col="orange")
legend("bottomright", legend=c("Empirical", "Binormal", "Density", "Log-normal",
                               "Log-concave density", "Smoothed log-concave density"),
       col=c("black", "green", "blue", "magenta", "brown", "orange"), lwd=2)

Improvements to the plotting function

Several users have been bothered by the fact than in pROC (R version), the sensitivity is plotted as decreasing specificity. Most other software plot increasing 1 – specificity on the X axis. The reason is purely historical: only few statistical software can plot an axis in decreasing direction. For instance S+ cannot do it, and pROC's ROC curve are plotted as 1 – specificity there. However it makes absolutely no difference on the ROC curve itself. As it was possible, I decided plot the modern version on R rather than stick to obsolete conventions.

For those who are disturbed and prefer to stick to obsolete conventions, pROC 1.5 comes with a way to plot increasing 1 – specificity in the R version with the legacy.axes argument.

plot(rocobj, legacy.axes=TRUE)

Note that it makes no difference to the coordinates of the plot, and if you want to add some text you still have to think in the "new" way. Consequently, the following will always be plotted to the top left corner of the curve, whatever legacy.axes you specified:

text(1, 1, auc(rocobj), adj=c(0, 1))

New return values in coords

The retargument of the coords function now accepts several new values:

• "accuracy": (sensivity + specificity) / 2
• "tn": true positives count
• "tp": true negatives count
• "fn": false negtives count (positive observations classified as negative by predictor)
• "fp": false positives count (negative observations classified as positive by predictor)
• "npv": negative predictive value, or tn / (tn + fn)
• "ppv": positive predictive value, or tp / (tp + fp)

In addition, sensitivity, specificity, npv and ppv can be prefixed with 1- in order to get the opposite value. Finally two additional values are recognized:

• "npe": converted to 1-npv
• "ppe": converted to 1-ppv

Here is an example. We take the best threshold of the ROC curve rocobj and display all the parameters of this threshold:

coords(rocobj, "best", ret=c("threshold", "specificity", "sensitivity", "accuracy", "tn", "tp", "fn", "fp", "npv", "ppv", "1-specificity", "1-sensitivity", "1-npv", "1-ppv"))

Coords also accepts a new argument: drop to control the dimension of the return value. If drop is FALSE, a matrix will always be returned, even if it contains only one column. This is especially useful to make scripts more reliable.

Conclusion

Here is the full change log:

• New cov and var functions
• coords accepts new ret values: "accuracy", "tn", "tp", "fn", "fp", "npv", "ppv", "1-specificity", "1-sensitivity", "1-npv", "1-ppv", "npe" and "ppe"
• New legacy.axes argument to plot 1-specificity rather than specificity
• New axes argument to turn off the plotting of the axis
• New logcondens and logcondens.smooth (Univariate Log-Concave Density Estimation) smoothing methods
• New function has.partial.auc to determine if an AUC is full or partial
• New argument drop for coords
• auc and multiclass.auc objects now also have secondary class numeric
• Updated load call
• Delong's CI reversed in ROC curves with direction=">"
• Delong's CI AUC returned values > 1 or < 0 in some rare cases
• Minor improvements in documentation

As usual, you can find the new version on ExPASy and on the CRAN. To update, type update.packages() or install.packages("pROC") if you want to update pROC only.

• 1. Elisabeth R. DeLong, David M. DeLong and Daniel L. Clarke-Pearson (1988) “Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach”. Biometrics 44, 837–845.
• 2. Lutz Duembgen, Kaspar Rufibach (2011) “logcondens: Computations Related to Univariate Log-Concave Density Estimation”. Journal of Statistical Software, 39, 1–28. URL: jstatsoft.org/v39/i06.

Incroyable ce qu'on peut faire avec HTML 5. Jérémie Patonnier présente un puzzle vidéo. Tout simplement bluffant !

I was invited to fill a profile on Careers 2.0. I received a bunch of invitations, if you want one contact me with your CV (as a link, or use the given email address) and I'll send you one.

For more than a year and a half, I have been releasing pROC version in two flavors: R and S+. The S+ version has a great advantage (or disadvantage for some) over R: it has a GUI, a graphical user interface. This is especially useful for users who do not want to learn programming R and who prefer a clic-and-point interface.

However, this dual-version comes with a cost. Here are some thoughts about it.

• I have to maintain two different version. Fortunately R and S+ are similar enough that most code works in both environments. There are a few inconsistencies, especially the way the variables are scoped (see this document by John Fox for more details). I need to be careful when sapplying: any variable accessed inside the function must be passed as argument in S+. Functions such as try and eval also behave quite differently. Overall, good tests in the examples section of the documentation are crucial to detect potential errors.
• The only exception is the plot function. Graphics have huge differences between R and S+. Any modification of plot is really difficult.
• S+ is much slower than R. Bootstrap operations take much longer to complete. Even DeLong computations, which are nearly instant in R, take a noticeable time to complete in S+.
• Specifically about the GUI now, writing S+ GUI code is awfully painful. For example there is little control over the placement of the buttons. The sample size window I'm currently developing is really ugly, and there is nothing I can do to improve it. The events are quite unpredictable, especially when the user rolls-back a window it is often in a broken status.
• The previous point stems from the lack documentation of the S+ GUI. It comes with a minimal introduction, that is just enough to start. But there are far too many undocumented features I had to discover myself by guessing-and-error. For instance, how to catch a user roll-back? I still haven't found how to cancel an analysis when I detect an error in the user's submission, or how to handle colors properly. For a professional-grade software, this lack of documentation is unexpected, to say the least.

So, would I start a new project with a GUI in S+? Probably not. I would first investigate the R integration plugin of SPSS. There are also some GUI for R, such as Rcmdr. I can't tell if they are a manageable option, as I have never tested it. Another option would be to create a web interface, but R doesn't have CGI and it would require interfacing with another language such as perl. Anyway, I would not recommend anybody to start an S+ GUI project.

What to expect of pROC in the future? Let me say I don't expect to stop its development soon. The version 1.5 of pROC, which will be released very soon, has a new sample size computation function, and a dedicated window in S+. In the long term (> 1 year) however, the situation might change. I will leave the University of Geneva in just over a year; will I still have access to S+? This problem is typical from proprietary software and its licensing. I won't pay for it, and I don't know if TIBCO provides licenses for application developers.

Dans une interview donnée à la BBC le 26 septembre dernier, Alessio Rastani déclare que ce ne sont pas les gouvernements qui dirigent le monde, mais Goldman Sachs.

L'analyste dit aussi que les plans de sauvetages ne changeront rien et ne l'intéressent pas. Ce qui l'intéresse, c'est comment gagner de l'argent. Il rêve depuis des années d'avoir une telle récession – pour gagner de l'argent. Aussi inquiétant que cela puisse paraître, sa prédiction vient de se réaliser. Avec Loukás Papadímos, la banque dirige maintenant l'Europe.

Alessio Rastani fait également une deuxième prédiction : dans 12 mois, les capitaux de milions de personnes vont disparaître. Voilà qui devient intéressant.

Nous sommes le 11.11.11. Dans quelques instants il sera 11:11:11. La dernière date binaire (qui s'écrit uniquement avec des 0 et des 1) avant un bon bout de temps (un peu moins de 89 ans, 88 ans, 50 jours, 12 heures, 48 minutes et 49 secondes pour être précis).

You can now follow pROC development on GitHub!

Among the upcoming features: sample size computation, variance and covariance functions, and much more… stay tuned!

I just released pROC version 1.4.4.

This version fixes a critical bug in the computation of one-tailed ROC tests p values. Please update to this version if you perform one-tailed tests.

library(pROC)
data(aSAH)
rocs100 <- roc(aSAH$outcome, aSAH$s100)
rocwfns <- roc(aSAH$outcome, aSAH$wfns)
roc.test(rocs100, rocwfns, alternative="greater")
roc.test(rocs100, rocwfns, alternative="less")

Thanks to Lisa Koch for pointing out the problem.

Here is the full change log:

• Fixed alternative for one-tailed tests.
• Removed COPYING file to fix a warning in r-devel.

As usual, you can find the new version on ExPASy and on the CRAN. To update, type update.packages() or install.packages("pROC") if you want to update pROC only.