Friday, February 11, 2011

Dropping the h-Bomb

Perhaps it was inevitable. Throughout our daughter's childhood, the teenage years loomed ever larger. We heard the stories, we knew what might happen. And then we got there and.. it was fine. In fact, everything has been great. Until a few days ago.

A few nights ago, during dinner, our daughter wanted to know the h-index of each of her parents.

What to do? She has asked us some difficult questions in the past, like when she wondered which parent is more famous (short answer: neither), but this question was somehow more.. personal.

We told her. My husband's h-index is higher than mine.

Will this affect how she views us? Should we have told her?

Is revealing our h-index a gateway to future nerdy questions? Will she now wonder how much grant money we each bring in? Will she start begging for chemical safety training, and then demand the keys to the lab?

I guess we will just have to keep doing what we've been doing: wing it. Despite a shocking lack of preparation for being parents in the first place, we will try to navigate the eddies and shoals of the teen years. Perhaps it is even time for us to start talking to our daughter about the importance of having an updated CV, but I'm not sure I'm ready to have that conversation yet.

30 comments:

  1. OhMyGod. I can see this in my future. My son's best friends include not one, not two, but three* boys whose parents are condensed matter physicists. Yes, all six of them.
    I can imagine harsh questions when he'll find out that to get a double digit h-index he has to sum his mum's and his dad's.
    There are times when I think being a research mathematician isn't so compatible with parenting.

    *That's actually true.

    ReplyDelete
  2. Absolutely a Gateway Question. I would start monitoring her PubMed searches and her activity on the NSF Award Search page. If she starts asking about overhead rates, you may need to bring in a professional.

    ReplyDelete
  3. Love it! We were unprepared for that question as well, as my in-laws solved it by publishing together. Would that we had thought ahead in that way.

    But I still remember when my now teen took my impact rating (much higher than my husband's) to school for show and tell when he was in 2nd grade -- a 3rd generation academic, what have we done?

    ReplyDelete
  4. Great Friday post!

    ReplyDelete
  5. So, here is a horrible situation. Our department has a habit of computing h- and g-indices after removing self-citations. However, we have one husband and wife couple whose work overlaps but who never publish together. This year, the promotions committee subtracted out his citations to her as well as self-citations when comparing her productivity to her cohort. This left her with a smaller raise than might have been had otherwise.

    Is this in anyway justifiable? (She is threatening to sue when it came out that the Dean had suggested this method of calculation.)

    ReplyDelete
  6. I'm afraid these may not be truly "nerdy" questions. You may have something even worse going on.

    I don't want to worry you prematurely, but it's possible that you may have a future dean.

    ReplyDelete
  7. Is this in anyway justifiable?

    What isn't justifiable is the use of the h-index metric for pay raises.

    The h-index can give you a general idea if a researcher is active or inactive, and if active how prominent, but it has no precision about small things such as year-to-year pay increases.

    ReplyDelete
  8. Anon 11:00: That is not justified. I hope the department also went through every faculty member's citation record in detail so that citations by close colleagues, friends, former students, and others with a potential conflict of interest can all be factored into the equation (which I agree with Anon 12:18 is not a good way to determine merit on an annual basis).

    ReplyDelete
  9. My father was a computer science professor. I have still clear memories of vocabulary surrounding his work which he spoke about around me starting when I was 4 years old. For example: grants, overhead, programming terminology, and every faculty member's last name. Very scary. I don't want my daughter to remember all my colleagues' names 30 years from now!

    ReplyDelete
  10. My parents are both physicists, and I'm not american by birth. The first phrase I learned how to say in english, when I was 6-7?

    Correlation functional

    Of course, it took some time before I got to the point where I knew what it meant... :)

    And Anon @11:00... you mean they did this for the wife but not the husband? It would be unfair even if it was done to both, but I'm wondering...

    ReplyDelete
  11. To damigiana: I'm a pure mathematician and my husband is a condensed matter physicist, I'll feel the pain of low h-index when my son grows up!

    ReplyDelete
  12. LOL..and I didn't knew about h-index till I entered graduate school :D

    ReplyDelete
  13. The H-index is a crap metric that illustrates much of what is wrong with academia today.

    Einstein had a crap H-index. So did Fred Sanger, who published fewer than forty papers in his career but won the Nobel Prize in Chemistry twice.

    @ Annonymous: if your department actually uses H and G indices to determine salaries... is it filled with dunderheads? So it would seem.

    *And yes, I do have a "good" H-index for someone at my career stage.

    **For certain, rotten values of "good".

    ReplyDelete
  14. When I was an innocent ten I once used a particular mildish swear word in front of my mum. She did not raise her eyes from her book but said firmly "you should not use words you don't know the meaning of". I was duly put in my place.

    Now whenever my son (8 yrs) uses any word (not rude words of course) I think he does not know the meaning of I challenge him to explain it to me. I also do this when we read books together at bed time. Most times he has to admit he does not know and so I use this as an opportunity to teach him....

    Now if my son asked about my H index I would have asked him what that was AND how is it worked out..... I am an academic of 15ys standing and knew diddly squat about H indices for 14.5 years of my career so far.

    My interpretation of H indices is that when it is "significant" it's meaningless as you would have known about that person's standing in the community via a multitude of other ways.

    ReplyDelete
  15. On the h-index:

    Einstein had an h-index of 57. This is far from crap, though admittedly below that of some other lesser famous physicists.

    Every metric short of auditing your entire academic record will be necessarily a mere approximation.

    The h-index is a clear improvement over the previous metric of straight paper counting. This can be a two sided thing: it is so much of an improvement that places which in the past refused to use straight paper counting now may use the h-index.

    The "criticism" that the h-index simply confirms what we thought---stated here as well as elsewhere--is most bizarre. Would one say that a weight scale is useless since it simply confirms what we knew already by picking up the object?

    Pointing out outliers is not enough to disqualify the h-index. Heck we are scientists here: every honest plot of an experimental measure has points where there shouldn't be any due to noise, equipment error, chance among other reasons.

    A large h-index is hard to forge. If you cite your spouse's papers or your friends' papers, that might get you at most the first 9 or 10 points in your h-index quest. To move a point beyond that from 10 to 11 up you need a minimum of 21 citations, from 15 to 16 you need at least 31 citations, from 25 to 26 you need 51 citations and an overall count of at least 676 citations for your top 26 papers. This is hard to fake

    In fact for an h-index of around 20 there is not even a need to weed out self-citations. By now it is so hard to move up that unless other people are widely citing your work you won't go any further.

    The big problems with the h-index are

    (1) it is age-specific,

    (2) it is (sub)field specific because of citation patterns,

    (3) it has a false level of resolution,

    This last leads to unsophisticated minds making distinctions between persons whose h-index differs by 1-3 points, which is not statistically significant.

    (4) it is susceptible to manipulation for low values through self citation and friend-citation agreements.

    ReplyDelete
  16. "Einstein had an h-index of 57. This is far from crap, though admittedly below that of some other lesser famous physicists."

    We have in our department multiple scientists with H-indices well in excess of 80, and in our institution at least a couple with H-indices approaching 150. No full professor in my department has an H-index of less than 30. Some very accomplished folks in this crowd, but I don't think that any of 'em have made a contribution comparable to that of Einstein or Sanger.

    I do think that the H-index imposes yet another set of perverse incentive in a system already dominated by them. It absolutely rewards publishing a lot of decent papers in preference to a few really spectacular ones — and it's not as though the system needs to nudge things ever further in that direction, further reinforcing the triumph of empire building versus scientific achievement.

    ReplyDelete
  17. We have in our department multiple scientists with H-indices well in excess of 80, and in our institution at least a couple with H-indices approaching 150.

    What was I saying about comparing h-indices across fields? In physics an h-index of 80 is very high. Here's a list of a few famous physicist selected at random:

    Richard Feynman* : 58
    Stephen Hawking: 78
    Juan Maldacena: 70
    George Smoot* : 44
    Roy Glauber* : 37
    Murray Gell-Man* : 48
    Vitaly Ginzburg* : 47
    David Politzer* : 47

    * Nobel Prize winners

    ReplyDelete
  18. It absolutely rewards publishing a lot of decent papers in preference to a few really spectacular ones

    That it does. However with very rare exceptions top notch researchers are both prolific and high impact scientists: if one is very active and undertaking many worthwhile lines of research there will be plenty of side-line and dead-end decent papers along the way to give you a high h-index.

    In the end, I agree. There is something to be said for the variants of the h-index (such as e-index and AR-index) which count the number of citations above the h-bar, but now we are talking about the details. This suggests that the h-index is not far from right. Think about the contrapositive: would we be quibbling about the details if the proposed scientist ranking method was, say, shoe size?

    ReplyDelete
  19. Hi FSP - I love Cordelia Fine's stuff - hope you enjoy this, and recommend her writing too.


    ttp://bps-research-digest.blogspot.com/2011/02/cordelia-fine-my-greed.html

    ReplyDelete
  20. The big elephant in the room with h-index is that it doesn't just ignore but in fact positively rewards honorary and bogus co-authorships. A single paper with 50 citations will add exactly one to your h-index (well it will to mine), whether you were sole author or 49th on the list. In the latter case, it also adds one to the h-index of all the other co-authors who also basically did bugger all. So sharing authorship with all your friends, and getting the same back from them, inflates everyone's h-index without any increase in actual output or citations.

    It would be trivial to correct for this, but no-one ever bothers, as far as I can tell. The guy who proposed the measure even agreed with me that this would make it fairer...

    ReplyDelete
  21. lol the talk of the h-index takes me back - it was one of the reasons I had to get out of biochem lol

    Cool blog - follwoing

    ReplyDelete
  22. Google Scholar's h-index computation has both regular and normalized for co-authorship indices. (I think that you need a plugin to get Google Scholar to compute h-index.)

    Co-authorship normalization does make a difference for me, particularly since many of my papers are joint with my grad students (and I usually give them first authorship):

    Impact indices:
    (Plain values)
    Citations selected: 4879 h-index: 29 g-index: 69 e-index: 61 delta-h: the number of citations needed to increment h-index by 1: 3 delta-g: the number of citations needed to increment g-index by 1: 21
    (Normalized The citations of each paper are divided by the corresponding number of authors (authors >4 are rounded to 4) per co-authorship)
    Citations selected: 2310.2 h-index: 22.0 g-index: 47.0 e-index: 39.0 delta-h: the number of citations needed to increment h-index by 1: 1.0 delta-g: the number of citations needed to increment g-index by 1: 14.2

    ReplyDelete
  23. in fact positively rewards honorary and bogus co-authorships. A single paper with 50 citations will add exactly one to your h-index (well it will to mine), whether you were sole author or 49th on the list.

    Yes it does, but again this applies only at low values of h.

    Let's consider a field where the average number of citations per paper is, say, 10. In such field, to get an h-index of 10 you need to work your way into ten of those honorary bogus authorships.

    Now to get an h-index of 20 you need to work your way into twenty above average papers.

    To get an index of 30 you need to work your way into 30 papers which are several standard variations over the mean.

    For a bogus h-index of 50 you need a scientist who can sweet talk its way into fifty well above average papers. That would be a major accomplishment all of its own.

    ReplyDelete
  24. "Co-authorship normalization does make a difference for me, particularly since many of my papers are joint with my grad students (and I usually give them first authorship)"

    Give? Give? *GIVE?*

    Look, I realize that that was a figure of speech, but words really do matter.

    I can't speak for you, but in my laboratory (mine in the sense that I'm the PI), the grad students and postdocs do the *vast* majority of the actual experimentation. It's more like *they* give *me* last authorship. I mainly write grants, provide advice that may or may not be of value, and edit/coauthor the manuscripts. These are not valueless services, but neither are they (for the most part) the heavy lifting.

    In general, one's contributions while a grad student or postdoc are grossly undervalued; one you're a PI, they're generally overvalued.

    Another reason why the H-index blows.

    In any case, first authorship that one should ever "give." If it's earned, it's not an award, and if it's unearned and awarded anyway, something is ethically amiss.

    ReplyDelete
  25. Re: "only is a problem at low values of h"

    Keep in mind that, as the other commenting mathematicians have already pointed out, in some fields, almost everyone has a single digit h-index.

    ReplyDelete
  26. Spiny Norman objected to my use of the word "give" for assigning first authorship to students.

    In some cases, his objection is well taken---the students have cleared earned the first authorship. In others, it is a gift, as almost all the ideas in the paper are mine and the students have done only a fraction of the work. There is no standard way to do "middle authorship" on a two-author paper, for example, so some incorrect assumptions about the amount of work done by the authors will be made no matter how the authorships are listed. I prefer to err on the side of being generous to co-authors, particularly students, who need first-author credit more than I do.

    I am aware of the difference between these cases, and I should have been more careful in my use of the verb. In most of my recent papers the first author is indeed the primary author.

    ReplyDelete
  27. Anon above,

    I'm not suggesting that a completely unproductive researcher would blag their way from nowhere to h=50, but that a moderate researcher can artificially inflate their h-index a bit, and the widespread adoption of its use actually encourages and rewards such behaviour.

    Given job adverts that either explicitly or implicitly say things like "h<20 need not apply", this is something that I think is worth paying attention to. A basic correction for co-authorship is trivial (and thanks to the person who pointed out the firefox addon that does this, but until people actually use this variant, it doesn't help much).

    ReplyDelete
  28. but that a moderate researcher can artificially inflate their h-index a bit,

    Any measure can be subverted. Even on a full auditing of your academic record you can go and bribe the judges. So just pointing out that you can cheat the measure says nothing.

    The question is how easy it is to do so?

    The h-index has the rather interesting property that the amount of effort required to manipulate it grows super-linearly. Compare this with citation counts and publication counts which are completely linear and as such much easier to subvert.

    ReplyDelete
  29. That is just hysterical, in a nerdy sort of way. What a perceptive child you have!

    ReplyDelete