AGU 2005 Draft

Update: a new heavily-edited version sent to me this a.m. by Ross, edited somewhat through the day here. Ross to the rescue.

Here’s a draft Powerpoint presentation for AGU. As with most first drafts, it’s pretty uneven. The topic of the session is variability, so I’ve spent the first part of the presentation on that. I haven’t even thought about time. I can delete sections at will. Comments are more than welcome – the more critical, the better ( don’t be shy, TCO). I’m a big believer in second drafts so I’m not too fussed if a first draft is rough. I know that there are some good points, but I’m probably trying to cover too much.

I’ll be leaving on Monday a.m., but need to finish by Sunday. I’m going to be having dinner/drinks with Huybers, Zorita (von Storch isn’t going) , Roger Pielke, hopefully some others out there, which will be fun. I’m going to take my squash equipment with me as my club has exchange with the University Club in San Francisco. If anyone here wants to contact me, I’ll be at the Westin St Francis.


  1. Armand MacMurray
    Posted Dec 2, 2005 at 6:42 PM | Permalink

    1) You need a big-picture outline slide at the beginning.
    2) Way too much text — you have 2x to 3x more text on most slides than anyone will be able to read/understand (except perhaps for faithful readers of this blog!).
    3) In principle, I like slides like 23 ( reconstruction), but it needs to be “repackaged” for impact. For example, something like:
    Q: “” stock prices: an excellent paleoclimate temperature proxy?
    A1: Some would say: The RE statistic is high, so it must be!
    A2: But the R2 is…[you get the idea]

    My advice: if you can’t say it using 24 point text, you’re cramming too much into a single slide.

  2. per
    Posted Dec 2, 2005 at 7:48 PM | Permalink

    you are covering a lot, and I don’t know what the audience is like in terms of whether they have any knowledge of what you are talking about; I frequently reduce the content of what I am giving so that I can make sure the audience understand what is at issue.

    You start off by introducing a problem, the variability problem. Is all the presentation about the problem you outline in slide 2 ? I wasn’t sure if the Durbin-watson stuff is addressing a different issue. Under any circumstances, where is the slide that summarises your answer to the question posed in slide 2 ?

    My overall impression is that this will be a very enjoyable talk; there is lots of interesting stuff in there, and the implications of what you are saying are really quite important. Should be a joy to give.

    best wishes

  3. Steve McIntyre
    Posted Dec 2, 2005 at 10:16 PM | Permalink

    I was thinking of the following for a big picture slide, but it seemed a little obscure even to me:

    1. What’s regression got to do, got to do with it?
    2. It ain’t the meat, it’s the method.
    3. Whatcha gon’ do with all that junk, all that junk?

    Points for spotting all three as this is multi-generational.

  4. Steve McIntyre
    Posted Dec 2, 2005 at 10:26 PM | Permalink

    Re #1: Thanks about the 24 point. I need to out my talking points from the slides.

  5. Posted Dec 2, 2005 at 10:31 PM | Permalink

    I agree, you need a big picture intro slide and then at the end of the presentation follow up with a closing summary that address the big picture points. Some slides have way too much text. Bullet points on the slide, put the words in your talk. Wish I could be a fly on the wall to see the audience response. Looks like a very interesting presentation. The best of luck!

  6. Steve McIntyre
    Posted Dec 2, 2005 at 11:40 PM | Permalink

    I’ve gone through and tidied a lot. Will post up 2nd draft tomorrow. I like writing point form conclusions, but sometimes you have to wait until you’ve finished to see what you’ve done. I think that I can summarize the talk into 3-4 points. If I can’t then, I’ll have to do a bigger edit than I want to do.

  7. Chris Chittleborough
    Posted Dec 3, 2005 at 1:22 AM | Permalink

    Re #3:
    1. What’s Love Got To Do With It
    — Tina Turner, 1984

    2. It Ain’t the Meat, It’s the Motion
    — Maria Muldaur, 1974 (had to google that one)

    3. My Humps
    — Black-Eyed Peas, 1998 (

    I think you’re right about being too obscure … 🙂

  8. Steve McIntyre
    Posted Dec 3, 2005 at 3:28 AM | Permalink

    Pretty good. I think that the last one is current: Black Eyed Peas: “My Humps” (Universal Int’l, 2005). It’s being played currently. I have the Maria Muldaur album – what was the lead song on it?

  9. Steve McIntyre
    Posted Dec 3, 2005 at 3:31 AM | Permalink

    Another trivia question: the Knot’s Landing reunion was on TV last night and was pretty ghastly; I surfed it since we used to watch Knot’s Landing a lot many years ago. Can you name three actors/actresses who appeared on Knot’s Landing who also won Oscars? Googling is a little unfair, but may not help with the third. Name another Knot’s Landing character who, in real life, was in a singing group that had a few big hits. Who are the two Desperate Housewives also appeared on Knot’s Landing? What is the classic Seinfeld phrase associated with Desperate Housewive’s Teri Hatcher?

  10. Bob K
    Posted Dec 3, 2005 at 7:42 AM | Permalink

    I’m not up on the math, so I checked readability.

    Slide #16 Durbin-Watson test. Double sp ace in the word “contains”.
    Slide #20 Cross-validtion R2 statistic. Meaning of first sentence is unclear.
    “In a spurious model, one expects to a reduction in the cross-validation R2 statistic in the verification period.”
    Maybe remove the word “to” or insert a verb following it?

    Keep up the good work.

  11. Willis Eschenbach
    Posted Dec 3, 2005 at 10:45 AM | Permalink

    Steve, it’s generally very good and interesting, but it needs an overall form. The form I usually use for this type of thing is:

    1. Tell people what you are about to tell them.

    2. Tell people the actual information itself.

    3. Tell people what you just told them.

    Presenting the same information three times, first in an abbreviated “here’s what you’re gonna get” form, then with all necessary details, and finally in a brief recap form, ensures that the message gets through.

    Best of luck at the presentation,


  12. mikep
    Posted Dec 3, 2005 at 10:54 AM | Permalink

    For some reason my powerpoint won’t open the draft – any ideas why not?

  13. John G. Bell
    Posted Dec 3, 2005 at 11:25 AM | Permalink

    Steve, Any way to get your presentation recorded? I would love to at least hear it. You could offer the video/audio on your web site. It would be a hit.

    Willis’s is right about the 3 parts to a talk like this. Wrap your talk in 1 and 3.

    Dang I wish I could hear it live.

    Have fun with it but get your points across. Great stuff you know, but have them take it home.

  14. Dano
    Posted Dec 3, 2005 at 2:19 PM | Permalink

    You should take a Tufte-type class, Steve. This PP is unviewable. What’s your thesis? What’s the point?

    Details: give the details in your handouts [available on the table at the back of the room] and stick to making slides on bullet points that you speak to (you’d better ensure you have enough, because folks are gonna take more than one to pass around, believe me). Ugh. Nothing is worse than a boring PP presentation and this is one of them. All that text is horrible. And where’s your conclusion. That stupid dot thing takes up too much room that you could use to make those likely dim and unviewable graphs clearer – make it a watermark. And Tahoma bold is more readable. Slides 13-14 s*ck – no one will read them. And the scales on the bottom of 12 likely won’t be readable on the screen due to projection issues, so you’d better ensure it is clear on your handout. And the background on 11 may wash out in the projection making some of the colored dots unviewable – greater variance. Graph on pg 2 won’t be readable in the back of the room, esp with that dot thing.

    And for chrissake use some refs – you look petulant in this presentation with no refs. Give the refs in your handout, not on the screen. You have a lot of work to do to not look bad in this presentation. If you, as your words claim, want to look credible, this won’t do it. Actions speak louder than words and you are yelling out your intent, and you can’t back it up in your presentation. You should honor people’s time by not wasting it.

    Get some help, quick, from someone in your Uni out of the Public school. This is unacceptable for professionals. Get input from people other than the Adoration Posse here. Drop the time-wasting trivia with the Adoration Posse and fix this thing.

    And I’m sure, JG Bell, that Heritage/AEI/CEI or somebody will video it – come now. Wake up.



  15. John G. Bell
    Posted Dec 3, 2005 at 3:19 PM | Permalink

    Dano, I’ll speeddial them up just to check :). You’re right I must have been asleep at the switch not to have done it earlier :).

  16. Gerald Machnee
    Posted Dec 3, 2005 at 3:49 PM | Permalink

    How much time do you have to talk? On average you need a minute per slide in order not to be rushed and have time for questions. Some slides have a lot of words but there is compensation in the shorter ones. If you have half an hour you can manage, If 20 minutes, you will be rushed.
    Anyway – have handout copies of the slides and indicate that every statement cannot be covered in the presentation – and copies are available for more detail. You want to continue the interest!!

  17. Brooks Hurd
    Posted Dec 3, 2005 at 4:20 PM | Permalink


    I have done way too many Power Point presentations. My problem is always to simplify each slide so that it contains just enough information to make the point clear. More words can hide the point. All the explanations can be in your talk.

    Here are some suggestions:

    Slide 7 – I would remove “step 1” and all the text after it.

    Slide 8 – same

    Slide 9 – same and reduce the large font text

    Slide 10 – make the text into several bullets

    Slide 12 – reduce the text to a couple bullets

    Slide 13 – reduce the text to the key point of each bullet.

    Slide 14 – same as 13

    Slide 15 – separate this into several slides and reduce the text in the second paragraph

    Slide 16 – leave the first sentence, the rest can be in the talk

    Slide 17 – leave the bullets and use the rest in the talk

    Slide 21 – Third bullet could read someting like “With correct cross validation R2 statistics confidence intervals are natural variability”

    Slide 23 – Leave the first sentence and cover the rest in the talk

    Slide 24 – The messsage is in the second bullet. You can cover the rest in the talk

    Slide 26 – You could shorten each sentence

    Slide 27 – Cover the third sentence in the talk – you might add something to the graphic to show the MWP and modern temps

  18. TCO
    Posted Dec 3, 2005 at 7:46 PM | Permalink

    Steve, it is not quite as ugly as the “poster”, but it is really a mess. Glad that you got something on paper (agree that that is important and it is fine to blather on a first draft). That said, you need to find a theme, organize content to fit your topic, then display it well. You have failed completely on the first two needs. Third, we could work on but why bother, when you don’t even know what you want to talk about.

    DanO is right in his comments. And he’s doing “his opposition” a clear kindness in telling you this.

    Go back, write something that you might actually present and then we can help you make that better.

    Having me write a 30 page description of every mistake here, might be educational here, but you can learn as much by just getting a book on how to write a science paper.

  19. TCO
    Posted Dec 3, 2005 at 7:47 PM | Permalink

    Or a book on how to give a science talk. there are quite readable ones that you can buy or get from the library.

  20. John G. Bell
    Posted Dec 3, 2005 at 8:02 PM | Permalink

    Dano’s idea about references in the pile of handouts by the exit is a must, repeat as many times as you want that they are available, but my idea was to leave them with something that they walk out the door thinking about. The best lectures I’ve heard have such things spun into the what I said part or the Q & A. You’ve got interesting stuff to talk about and know it insideout. Hope you get good questions. Bring a few with you to help get things going.

  21. Steve McIntyre
    Posted Dec 3, 2005 at 8:59 PM | Permalink

    Thanks for the various comments. The trouble with working up new material is that it’s uneven. Anyway I’ve trimmed it back a lot. Some wise people have told me to use 15 slides for 15 minutes. Added to everything else is that it’s been hard to be on form this week.

  22. Steve McIntyre
    Posted Dec 3, 2005 at 9:13 PM | Permalink

    I’ll sort this out. I don’t disagee with the comments. I wouldn’t normally have posted something this rough up, but I’m frazzled for obvious reasons, behind schedule and trying to force feed this.

    Part of the problem in deciding what to say is trying to be responsive to the session theme – variability – which is not a primary concern of mine, but does not harm to think through.

    It’s a bit of a risk trying out new material and seeing what works.

  23. James Lane
    Posted Dec 4, 2005 at 1:18 AM | Permalink

    Hi Steve,

    Coming in late with this, but basically agree with the comments above. Most importantly:

    1. A slide to say what you’re going to say

    2. Slides to say it

    3. A slide to summarize what you’ve said

    I do hundreds of PP presentations.

    Mostly, I think of bullet points as reminders (for me) of what I’m going to say. As such, they are very brief. Any other way, the audience are going to try to “read” the presentation and won’t listen to what you’re saying. All the detail and references can go in the handout.

    Not sure what the timing is, but if you want to email me a revised version, I’ll try to make prompt comments. Good luck with the presentation.

  24. Steve McIntyre
    Posted Dec 4, 2005 at 7:11 AM | Permalink

    I’ve put in a new draft as sent back to me by Ross, who has far more experience with academic presentations than me. Ross to the rescue once again. I’ll nuance some of the points a little, but I’ll go with something pretty close to this. It’s always a little painful losing details that I like, but been there, done that.

    The sales-orientation of academic presentations amazes me. Nothing wrong with that, just learning the lay of the land.

    One thing I like about blogging is that it’s possible to spend time on some detailed points. I appreciate the fact that there’s an audience for this, since I’m firmly of the view that the devil is in the details.

  25. TCO
    Posted Dec 4, 2005 at 10:06 AM | Permalink

    I don’t think you should be presenting new work at a meeting like this. An informal coffee or this blog is one thing. but not a big meeting like this. you should at least have worked it out and written a detailed post about it here (if not actually published).

    It’s also unclear to me what your topic is: examining Stork’s hypothesis (and is his hypothesis the first para or second–neither is a clear statement…why don’t YOU clearly label a simple hypothesis? An overview on climate variability by paleoclimatology reconstruction? Faults in same? Just various errors of your “opponents”? I think you are much too broad in your topic.

    If you really want to give broad overviews on the subject overall, you need to step back and say that you are doing so specifically (and you won’t be able to present much new micro level experimental findings in such a talk…just main points at a topical level). Also, you should have a wealth of published work to your credit to justify such a standing. In fact you probably have spent the requisite time immersed in the field’s literature to have your say…however, you need to get some notches on your belt too…clear writing is clear thinking. You’re deliquent in writing good journal articles. And if they are anything like the two examples, we’ve seen here (poster and AGU talk), then it’s no surprise that you’re having issues getting published.

    Will take a look at new version on a computer that has PP.

  26. TCO
    Posted Dec 4, 2005 at 10:09 AM | Permalink

    Oh…and I agree about all the sales stuff in science papers. Many of them are written to get grants (frigging rat race it becomes). That said, your problem is not lack of sales stuff or such. It’s just not clear writing.

  27. Steve McIntyre
    Posted Dec 4, 2005 at 11:12 AM | Permalink

    Other than the Nature situation, Ross and I haven’t had difficulty getting published. It’s much better doing stuff with a coauthor since you hash stuff out. I think that our actual articles read pretty well.

    My read of the AGU conference is that people present new material. I don’t think that there’s much of a market for re-hashing MBH one more time. Last year, I presented our MBH material, which was then new. The session topic was not really ideal for my interests. I probably should have just submitted something on Jones et al 1998 that I’m totally familiar with and taken my chances on being accepted.

    I took my chances on having some new material ready, which may or my not prove a good decision, but I have to live with it. In part, I wanted to force myself to work some material up; in part, I suppose that I was reacting to some claims that were being rpesented that the VS inverse regression hypothesis was the “real” thing that broke the hockey stick. While I appreciated VS public cordiality, we were also rivals to some degree in this microscopic enterprise and I didn’t think that his inverse regression hypothesis really had anything to do with the variability issue. So I wanted to comment on it.

    As it turns out, there’s hardly anything to say about the VS inverse regression hypothesis. It doesn’t begin to hold water. But climate scientists were so anxious to deny us any credit, that they were quick to over-praise the VS critique.

    So what am I really trying to do here now that it’s 5 minutes to midnight. Here’s a try at a new abstract, which seems reasonably coherent to me:

    Several explanations have been provided for differing low-frequency ranges of millennial climate reconstructions: proxy geography; inverse regression and “non-conservative” tree ring standardization. I argue here that none of these explanations is valid, and that the low-frequency ranges are directly linked to problems in calculating variances of highly autocorrelated series on short segments. These estimation difficulties are a symptom of more fundamental modeling problems in the canonical reconstructions, which are typically marked by adverse calibration period Durbin-Watson statistics and very poor out-of-sample R2 statistics. The similarity in performance in many reconstructions appears to be due to overlap of critical proxies.

    Does that tie things together any better?

  28. Armand MacMurray
    Posted Dec 4, 2005 at 12:28 PM | Permalink

    Wow – the new abstract is much better! I wonder if you could modify the wording of the last sentence to emphasize the common-proxy issue a bit more (I think “shared” is better than “overlap” in describing what’s going on). Perhaps something like “The similarity in performance among many reconstructions appears to be due to the shared use of certain critical proxies.”

  29. Armand MacMurray
    Posted Dec 4, 2005 at 12:44 PM | Permalink

    Powerpoint looks pretty good, too. Some minor issues in it:

    Slide 5 ((i) Differing Geographical Coverage): the 2nd and 3rd bullets appear to be points *you* are making, even though they are really what *others* are arguing. Could you start each with “They claim that…” or “They say that…”?

    Slide 7 ((ii) Inverse Regression Hypothesis): again, these are VS’ arguments, but the 2nd and 3rd appear as though *you* are making them.

    Slides 11/12: The first is titled “Alternative Explanation”, but the bullet items on that slide are just background/support for the real alternative explanation, on slide 12.

    Slide 17 (“honest” confidence intervals): It would be nice to give an idea of the relative size of the “honest” intervals vs. the commonly published ones. Are they 50% larger? 2x? 10x? 100x?

    Slide 19 (summary): The last point about verifying out-of-sample performance is important, but didn’t appear in the previous slides (I think!).

  30. Steve McIntyre
    Posted Dec 4, 2005 at 1:33 PM | Permalink

    Thanks Armand: changes made to 5,7,title of 11. Changed title of 19 to Conclusions from Summary. I don’t want to make an essentially editorial point twice in two slides. As to 17, I haven’t figured this out yet. As so often, it’s easier to spot a problem than to fix it. As a guess, I would say that there’s at least a 75% chance that the true confidence intervals from the reconstruction itself (expressed purely statistically) are larger than natural variability so that there is no net gain from the reconstruction itself. In an interesting article in Holocene a few years ago, David Lucy argued that you needed a really good R2 (which is not observed here) to reduce confidence intervals below natural variability. I corresponded with him about a year ago and he warned me in a very kindly way that, if I had any aspirations to academic advancement, I should stay away from trying to criticize climate reconstructions (due to their sacred character rather than their inherent merits). Since I had no such worries, he wished me luck.

  31. Neil Fisher
    Posted Dec 4, 2005 at 3:37 PM | Permalink

    if you get the time to do it, a “dry run” of your talk with people specifically instructed to “play dumb” about the subject at hand is very valuable. Not only does it help you fine-tune the timing of your talk, it also brings things to light that others here have already pointed out, such as too much info. per slide, extra hand-out stuff and the like.


  32. Armand MacMurray
    Posted Dec 4, 2005 at 3:55 PM | Permalink

    Steve, re #30, where you say

    As a guess, I would say that there’s at least a 75% chance that the true confidence intervals from the reconstruction itself (expressed purely statistically) are larger than natural variability so that there is no net gain from the reconstruction itself.

    I will pull on my TCO hat and *strongly* encourage you to write a real paper about this. Not only would it be easy to understand the overall gist (and broader relevance) for non-stats people, but the content would be almost all stats and thus well-suited to your skills. It bothers me that the reconstruction field seems to blithely ignore the need for useful error estimates (that is, estimates that should be valid for the reconstructed temperatures and not just for some technical subset of the process used to derive the temps).
    Of course, you are allowed to wait until after AGU and after things settle down a bit with your family. 🙂

  33. Armand MacMurray
    Posted Dec 4, 2005 at 3:56 PM | Permalink

    PS – Good luck & if they do record the talks, please post links.

  34. Steve McIntyre
    Posted Dec 4, 2005 at 5:35 PM | Permalink

    Although my methods differ, the idea of comparing statistical confidence intervals to natural variability is expressed clearly in Aykroyd et al. here

    the average 95% confidence width of the estimate can in some cases exceed the variability of the parameter being estimated, if the correlation between proxy and climate variable is relatively low.

  35. per
    Posted Dec 4, 2005 at 6:33 PM | Permalink

    looks great.

    couldn’t help but wonder if #18 is appropriate; it seems to be tangential to the main points that you are making. The conlusions slide #19 has some fairly important points in it, and I think the out-of-sample issue you raise in the last point is very important, and fits in with your previous points.
    good luck

  36. John G. Bell
    Posted Dec 4, 2005 at 7:32 PM | Permalink


    Text in your bullets on 9, 11,17 & 19 seems a bit large. They sort of flow together into unintended paragraph. Touch smaller text for them with a bit more space between bullets makes them more readable to me.

    Don’t they have December 5th on them? Did you want that?

    Best, John

  37. John G. Bell
    Posted Dec 4, 2005 at 7:37 PM | Permalink

    Uhm, the 5th is right. My watch is off. You would know :).

  38. Steve McIntyre
    Posted Dec 4, 2005 at 8:04 PM | Permalink

    Thanks to everyone and especially Ross. I’ve incorporated most suggestions. I’m leaving slide #18 in for now, but will be fast on the trigger.

  39. Paul
    Posted Dec 4, 2005 at 8:56 PM | Permalink

    #2 – Last Bullet

    Currently reads: …which are typically marked by adverse calibration period Durbin-Watson statistics and very poor out-of-sample R2 statistics.

    Should it read: …which are typically marked by adverse calibration period, Durbin-Watson statistics and very poor out-of-sample R2 statistics.

    Note the comma after period. Is this correct, or am I reading it wrong?

  40. Steve McIntyre
    Posted Dec 4, 2005 at 9:37 PM | Permalink

    It’s correct as written – DW stats in the calibration period.

  41. TCO
    Posted Dec 5, 2005 at 7:13 AM | Permalink

    1. I would argue that any of the reasonably well thought out, but not published (yet) arguments on this board. I.e. anything other than MBH criticism is “new”. You are creating two too-far-apart extremes with new analysis (that you are actually crunching now) versus the MBH slams.

    2. Your abstract does pull together and create a theme for all your slides. I still wonder if you are making too broad of a theme, though. And if you really know what your point is. Is it: “bad statistical methods in paleoclimatology”, “why do reconstructions differ from each other”, “why are they all wrong” or what? Actually, I think that you need to do a discrete paper on autocorrelation. You always mention it and such. But it is worth of individual attention. Honing in solely on this issue. (At times, you have an annoying almost warmer tendancy to throw the kitchen sink at things and to confuse ripping a paper to shreds with examining a specific issue.)

    3. I will look at the new PP though.

  42. Dano
    Posted Dec 5, 2005 at 7:36 PM | Permalink

    Next you go out and collect some data.

    This (now viewable) .ppt will at least get people to look up to see what’s causing the dogs to bark, but if you don’t show people what’s in the tree, this will be just another buncha yippy-*ss little dogs, tugging on your pant cuff, waiting to be a field goal.



  43. nanny_govt_sucks
    Posted Dec 5, 2005 at 9:05 PM | Permalink

    Dano, is it so hard to understand that some people are good at cutting up trees and others are good at math?

  44. Armand MacMurray
    Posted Dec 8, 2005 at 3:06 PM | Permalink

    Re: #43
    I’m sure you would agree that the FDA doesn’t have to go out and find a cure/treatment for a disease just so they can rule that another drug submitted to them doesn’t work or isn’t safe enough.

  45. Mats Holmstrom
    Posted Dec 9, 2005 at 9:34 PM | Permalink

    Just to let you know,
    Steve’s talk today was very good. I think he got the main points through, without rushing it. He said that it was the first time he had an oral presentation for an academic audience, but that did not show.

    There was 60+ people in the audience, and the reception of the talk seemed good.


  46. John G. Bell
    Posted Dec 9, 2005 at 10:17 PM | Permalink

    How were the questions? Was there a good exchange? Did they seem to get it, that is, to really follow the material?

  47. Mats Holmstrom
    Posted Dec 9, 2005 at 10:50 PM | Permalink

    Hi John,
    there was two questions, but I think I will let Steve write that up. It is not my field, and I do not want to misrepresent anything. Anyway, the questions were constructive, but AGU conferences with 10-15 min talks does not allow much discussions. Probably there was a lot of talk at the coffe break afterward, but I had left by then.

    Regarding science: It seemed to me that Steve’s concerns were taken seriously. If anything happens is another thing, but I have been to several controversial presentations where the insults have flown, and this was not one of them.

    In fact, my impression, as an outsider, was that people was genuinely interested in Steve’s results. I only attended a couple of talks in the session, but all seemed to have a critical view of data. So maybe Steve have started something. This might also be the first example of “blog science”…


%d bloggers like this: