jump to navigation

An appetizer for the impatient lubologist November 19, 2008

Posted by dorigo in internet, personal, physics, science.

I cannot resist directing the Lubophilites and the Lubophobes among you to the thread of the former post, where, out of the blue, Lubos starts an attempt at explaining why Strassler’s estimate of the cross section of “ghost events” in the recent CDF publication is right, and I am wrong.

Of course, I am right and Strassler is wrong. I know it because I authored the paper, but even if I had been so careless as to not know what I signed, I did talk about that very number -the cross section error of Strassler- with the main author of the CDF study. In any case, Lubos in the thread shows just how arrogant he is and -to his credit- how much he believes in himself, launching himself with a smile in the den of the tiger. Indeed, I explain to him what his error is (he can be excused for that: the CDF publication is not too clear about the fact that most of the study is performed with 742/pb, and only a part with 2100/pb), but he ignores my warnings, and ends up in a really sorry situation: having to choose between two evils. Defend his mistake ad infinitum, showing the world he is childish beyond repair, or retreat in good order ?

Of course, if you know Lubos, you know what he will do: and in fact, he counterattacks, ending even deeper in trouble. Some of the sentences of his last comment show just how deep his ignorance is.

An appetizer of better things to come -I will have a detailed post out tomorrow- below, pasted from his last comment (oh, as of now… He might please us with others):

The number 742/pb only tells us that in about 1/3 of the proton-antiproton collisions, a bottom-antibottom pair was believed to be created at the very beginning, as evidenced by the dimuon trigger. One could use this number, 742/pb, if it were true that all these events recorded with the dimuon trigger had the bottom-antibottom pair at the beginning of the event, and if all the cross sections were somewhat re-expressed as cross sections for bottom-antibottom collisions. Well, except that the bottom-antibottom are not really colliding here. They are just produced in 1/3 of the collisions.

1/3 of the collisions ? OMG LUBOOOOOOS! :))))

Now this post will appear as a “revenge” against Lubos Motl’s last post, where he uses a pair of pages to bitterly criticize me for my post on Strassler. I deny that: I in fact had posted on his blog the following comment (I paste it here because he might remove it, if  I know him…):

Nice article. Very wrong, but quite readable.

The authors of the CDF paper on multi-muons are not a unknown subset, since they all signed, and their names are on the front page. What is unknown is the fraction that did not sign, because authorship in CDF varies from paper to paper, due to people leaving CDF, new members, visitors who sign only one paper, etc.

About the rest… [Invalid characters removed.]


As you see, nothing aggressive on my part, and the exchange happened before the thread in my blog turned bad. But of course, I was not fast enough – after finally understanding his blunder, he removed the part where I explained his mistake, lest his readers understand he is wrong, and added a copy of his “explanations”. Oh well. Everybody is entitled to their opinion, even on Lubos Motl.


1. sciencetourist - November 20, 2008

I don’t have a professional level of knowledge in these sciences. I read trying to learn some stuff outside of my field and enjoy reading a few science blogs especially when they are about science. I keep bumping into this Motl.

My question to those in the know: Does he do any good/real work or is he just an asshole?

2. Luboš Motl - November 20, 2008

Dear Tommaso,

my pedagogical patience is virtually unlimited and my experience with dealing with slow and aggressive people is vast, so let me give you one more chance.

You claim that the 153,895 ghost events were chosen from the 742/pb sample, right? Open the CDF paper

Click to access 0810.5357v2.pdf

on page 16. Table II shows that 143,743 events passed the tight SVX filter. None of them was a ghost event, the same second column says. Now, open page 12. Table I claims there to summarize the events in the 742/pb sample (see the caption). If you add the entries, you get exactly those 143,000+ events, i.e. exactly the collection that included no ghost events.

To summarize these two tables I,II, there were no ghost events in the 742/pb ensemble. You must go to the SVX-unconstrained larger set of 2100/pb events to see the 153,000+ ghost events. Among the loosely SVX-filtered events, corresponding to about 1400/pb, Table II also explains that there were about 72,553 events.

Now, my predictions whether you will remain unable to get this simple point now are ambiguous because you would almost certainly become the densest physicist I have met in my life – trumping even people like Lee Smolin – if you don’t get it.


3. carlbrannen - November 20, 2008

Tommaso, the phrase “in the hole of the tiger” might promote a less than delicate understanding in the native English speaker’s mind. Perhaps “den of the tiger” would be a better phrase. Or perhaps you meant this exactly as you wrote it.

4. Luboš Motl - November 20, 2008

Incidentally, it is also a lie that I started to explain you these basic things about 75 pb “out of the blue”. I explained them to you because you politely asked me to do so. You wrote:

“In any case, I make a few points, and on those I can be criticized. For instance, please explain the 75 pb estimate of the paper.”

So I respected your wish and explained you what’s wrong with your calculation of the non-75-pb cross section.

5. Dr B D O Adams - November 20, 2008

On personality of Physicists. Of course Physics need be believe in the
intellegence, and the rightness of the theories, be able to defend them.
There is a need to recognise, when so there own theories is wrong. Its as if physicist need to switch between a political mode, for arguing there theories a right, and creative mode to creating new ones. Is having gained respect, and friends, at the end of each cycle is then, then the personal win? or is it having the rightness of there theories argued on the personal win? Finally, a wrong theory is agreed upon as right is a lost to the whole of science. Mathematical proof and evidence wins, we hope.

6. dorigo - November 20, 2008

Sciencetourist, Lubos is … well, Lubos. We have to live with the fact that he is there. The world would be a different place without him.

Lubos #2, I gather you are now slowly understanding where your mistake has been. It must hurt to see that I had pointed it out from the very beginning, but you were blind in your furor. Now, I think you should remove your last post, it adds too much color to your mistake.

Carl, thank you – will change the text (in a sec).

Lubos #4, I agree, you are right on that one. I tricked you into explaining the wrong figure by Strassler, since you defended his paper so actively.

Cheers all,

7. carlbrannen - November 20, 2008

Of course these things are confusing to me, but I think Lubos’ comment needs to be more fully critiqued by Tommaso in order for us to understand why the paper does not say what he claims.

Let me try and start this.

Regarding Lubos’ note #2, the caption for Table I reads “Number of events attributed to the different dimuon sources”, the important adjective being “attributed to”. So that it should sum to the figure in Table II is not a surprise and does not indicate that the 2100/pb was used as the source for the ghost particles.

8. dorigo - November 20, 2008

Carl, I am sorry but I am done commenting every line of an incompetent writer. I explained it rather clearly in various occasions in this blog. Let me try one last time:

153000/2100=75: this is what Strassler must have computed, and what Lubos subscribes to.

153000/742 =200 (not 75): this is what I pointed out.

Now, is the 153000 event sample obtained by filtering (with suitable cuts which have no effect on the “effective cross section” 153k events correspond to) 742/pb or 2100/pb ? That is the heart of the question.

It is 742/pb. It is not 2100/pb. It is made clear only at the bottom of page 28 of the CDF paper. Indeed, as I wrote elsewhere, the CDF publication is not very clear on this.

So why would CDF publish an analysis of such an important signal, and only deal with a third of the available data for cross section estimates and event counts ?

Because the later data are collected by a trigger which prescales the events. It collects only every second, or every fifth, or every tenth event, depending on how large the proton and antiproton bunches are at the moment. This is required to avoid flooding the data acquisition bandwidth when the instantaneous luminosity is very high.

When dealing with prescaled data, cross section estimates are harder. So Giromini et al. use unprescaled data to evidence an excess and size it up, and use all data together to study the kinematics.


9. Anonymous (no, I'm not Lubos Motl) - November 20, 2008

Dorigo, let me give you a suggestion for improvement.

You style is very different than Motl’s: he tries to explain every step of the argument, while you repeat the same formulas over and over again. Even if I had no clue of what was being discussed, it’d be a lot easier for me to believe in him. And I’m sure that’s why his blog is so popular even among non-physicists.

That said, realize that there are no ghosts in the 742/pb set, that you made a mistake, and you’ll gain everyone’s respect.

10. dorigo - November 20, 2008

Anon, whoever you are, you are mistaken in many ways. First of all, you really seem to have no clue, because if you had, you would ask Lubos to retreat, not me.
Second, because there is no need to explain many things. It really is that simple: the ghost contribution to CDF data exists regardless of how much luminosity you consider. Do not confuse luminosity with datasets with different cuts!

The 153000 events belong to a dataset collected initially, which is better controlled (no dynamical prescale). In it, the cross sections can be computed with smaller systematic uncertainties. The luminosity of this dataset is 742/pb.

Then, a larger set (which includes the former) is used for kinematic studies. It amounts to 2100/pb.

The arrogance of Lubos, who tries to teach an author what he wrote on a paper, might be a healthy thing, unless it becomes ridiculous.
I said it before: what I am writing is endorsed by the person who did the analysis. Feel free to contact him at frascati (at) fnal.gov , his name is Paolo Giromini. Say I asked you to contact him.


11. Dan Riley - November 20, 2008

In HEP, integrated luminosity always refers to the luminosity of the underlying data set as collected at the detector (ignoring for the moment subtleties like trigger prescaling). We don’t generally talk about the integrated luminosity of a subset of events, unless that subset corresponds to all the events for a well defined period of data collection (e.g., all the data for an hour, week or month of running). The idea that changing the filtering criteria could change the integrated luminosity is incompatible with the way the term is used in HEP. We don’t assign luminositites to filtered selections of events that way, as the luminosity represented by such a subset it isn’t a particularly useful or well measured number.

The columns of table II are alternative analyses of the same set of data, the results of applying different selection criteria to that data. The underlying data set that is the input to all the analyses in table II has an integrated luminosity of 742/pb, hence for all the columns in that table, the integrated luminosity is 742/pb. Those analyses did not look at data beyond the 742/pb set, so there is no way that anything in that table can represent more than 742/pb of data.

Lubos’s idea that different filtering criteria correspond to different integrated luminosities is incompatible with the standard usage of those terms in HEP. It’s really that simple. This isn’t a question of tricky math or anything, it’s a matter of the conventions of the field, and Tommaso is quite evidently better versed in those conventions than Lubos.

12. Saving a good text from a few mistakes « A Quantum Diaries Survivor - November 21, 2008

[…] isn’t it ? This is very, very wrong, as a reader, Dan Riley, well explains in a thread here. HEP experimentalists do not do that: they do not assign integrated luminosity to […]

13. dorigo - November 21, 2008

Hi Dan,
your explanation of Lubos’ mistake is very clear, and in fact I have linked it from a post where I correct a post by him, see here.


14. fisica delle particelle: scoperta e reazioni della blogosfera | Mondoingegneri - November 26, 2008

[…] é scoppiato nei rispettivi blog piu in altri un acceso dibattito, che a mio avviso, al di lá dei toni a volte un po’ aspri e accesi,  ben dimostra come il blog possa essere  un ottimo strumento di dibattito […]

Sorry comments are closed for this entry

%d bloggers like this: