Comments from Aart: =================== Dear all, Here are my comments and questions for the PDF draft (sept 25). best regards, Aart. % thanks 26: Is Ref [9] is outdated? (the 1/fb result is now 'to be published in prl). (I did not check any outer refs). % fixed 27+ This paragraph is the place to clearly explain how this paper differs from the previous one and why it is interesting. I feel thiscould be improved. e.g.: 'In this paper, we represent an update of the CDF analysis [7,8] with double the data sample (1.9/fb) and improved analysis techniques. The acceptance has been increased by.... etc' % thanks, rephased 34: Was 'discovery potential' (rather than expected limit) really studied? % the sentence is removed, but the expected limit is about 2 sigma discovery potential. In this case, they are the same. 39: 'for the first time': this looks like you're claiming you're the first to ever use a NN. Also 'event discriminant technique' doesn't mean much to me... % removed 'for the first time' and change 'event discriminant technique' to ' multivariate discriminant technique' 41-42: This information belong in the previous paragraph. % fixed 110-112: Suggest to move to top of Sec III % fixed 117-119: Isn't this cut inefficient if you have additional hard ppbar interactions? Why not require z(bjets)-z(lepton)<5 cm. Why the detour using the primary (from which the b-jets will be partially excluded)? % The primary event vertex is selected based on the class 12 vertex with highest sum pt of tracks. So, it's un-likely impacted by additional interactions. 124: maybe remove 'high-pT' % keep it as is. 125-128: could be shortened, eg: "The selection criteria are the same as for the central region, except for the \met requirement." % fixed 129-134: Since the plug region is new in the paper, perhaps it would be nice to give the reader some more handles to understand what's done here. Perhaps a figure before and after the new \met cut so show that indeed this complicated cut helps? As it is now, it's hard for the reader to conclude much beyond: "they did some complicated cuts and they say it helps.". % rephased and it should be more clear. 133: don't understand delta phi: what does 'the angle between ... /each/ jet' mean? % no longer needed 143 (and lots of other places): It would be good to use consistent notation for the 3 tagger categories throughout the paper. % thanks, will fix. 149+ Section IV seems to be made up out of several stand-alone pieces of text. I think it could do with some streamlining and removal of duplication. So first a general introduction of what a b-jet looks like and then, for each algorithm, a description of how they try to look for that signatures. For example: 160-163 is general and should go to the introduction of section IV. 156-159 seems more appropriate at the start of the NN-tagger section % reorganized the section and hope it reads better now. general remark: While I appreciate the detailed description of the b-taggers, I feel does cause some imballance to the paper. For example: virtually nothing is said about (b-)jet energy measurement, which is also quite important for the performance. % That's a good question. Unfortunately, the b-jet correction is still not available at this time. 228-234: I think this is a bit long and has some ugly jargon ('cut value'). How about 'Events are considered tagged if NNb-l>0.182 and NNb-c > 0.242. These criteria were choosen to be 90% efficient for SECVTX tagged b-jets. They reject 65 (50)% of the light (charm) jets tagged by SECVTX.' (and move the sentence about the 97% SF to the previous paragraph, where you explain the measurement of this SF). % rephased Table I: I don't like the 'SECVTX-independent'. I anyway see no need to make the distinction. % Table I is removed now 298 Non-W QCD Multijet - a figure w/ the ABCD etc regions would be nice. % rephased, 340 - It would be nice to give some numbers that lead to the 25% - Throughout, systematics seem to address uncertainty on the rates. What about the shapes (of the final NN-discrimant)? % The shape systematic is small and ignored so far. Added a sentence in the text 373-376 not sure I understand % rephased (10) when written like this, it seems to imply that eff_trigger and eff_kinematics are assumed to be independent. I assume eff_kinimeatics is with triggered events in the denominator? % These efficiencies are defined in sequence. 456-457 something is missing inbetween these lines. % fixed 489: what is the bg model used in the while training the NN and while optimizing the input-parameter choice? % the background consists of equal amount of wbb, ttbar and single top. added a sentence for that in the text. 466: I don't understand why you put Br(W->lv) in the effiency. % it's acceptance, not just efficiency 526-529: This should be swapped to indicate that the NN distribution could also have yielded a signal: 'We use a binned ...Figs 6-7 to test for the pressence of a WH signal. No excess over the background is observed. We therefore poceed to set a limit'. % fixed 531: You don't really need this equation as it already in (13). % removed eq (13) 549: Don't you need shape systematics? (for example for the non-W bg) % again, the shape systematic is small. Figures: Figures could be improved: In b/w W+HF and nonW look the same; - Higgs signal is often small and hard to see. - Fig 6 &7 could use the space better (i.e. zoom on the y-axis). - Fig 4 & 5 need units on the x-axis. % Ok, will remake the plots - I think it would help if figures could be grouped by quantity plotted or by tagging category. I.e. I'd suggest combining: fig 1 and 2 fig 4 and 5 fig 6 and 7 % will combined two double tags together. - Why not show figs 4&5 for the 1tag category, where you have a factor 10 more statistics? % Not sure there are needs for figs 4&5, removed.