Tuesday, December 30, 2008

Enhancing research in IISc

My colleague,Abi, has provided a short writeup on my another colleague's (Jayant Haritsa) talk in Pan-IIT on enhancing research capabilities in IIT/IISc.

Jayant makes very valid points in his talk and it is always enlightening to listen to him. He rightly remarks that we should choose one criteria, stick with it and try to excel in it. Without excelling in anything, we try to become all rounders. I am reminded of Irfan Pathan, who was a good bowler who was asked to open the innings and that completely destroyed him both as a bowler and batsman.

In his talk, Jayant mentions that the CS department publishes around 0.1 papers in Tier I conferences per "smart person" per year per faculty and mentions
At IISc now, quite a few departments, especially in the Electrical Sciences division, have authored documents outlining their expectations and modes of evaluation for both faculty and students. As a simple case in point, a recently implemented policy in the Computer Science department is that ME students can be considered for an S (outstanding) grade in their project work only if they have submitted a paper to a journal or conference that is internationally recognized to be of A+ or A calibre.
I did an analysis of the number of papers published by M.E students in all departments in IISc. For this, I took the names of the M.E. students who graduated in 2006 along with their department affiliation and searched in the Web of Science for publications. The reason I took 2006 and not 2007 or 2008 was because I assumed that they would not be any more publications from this group. I noticed that less than 10% of the M.E. students had any publication (While Jayant talks about conferences, I am talking about journal publications). In fact, I am including students who had coauthored papers with other research students in the group. In fact, the 20-80 principle does not hold with 8% of M.E students contributing 92% of the papers published.

In his talk, he mentions that the above policy has resulted in an increase in the number of publications. I would be interested to know whether the 20-80 criteria for research publications in IISc has been breached by this recent policy change.

After reading this, one may get the impression that several departments have followed this wise course but it is restricted to two departments, while there are 42 departments in IISc. Many departments in IISc would be shocked if criteria like publications or citations are used to judge students for grades (or even faculty for promotion/awards). In fact, in many departments, both the faculty and student are more likely to be discouraged if a M.E. student publishes in Tier IA journal in the field. (S)he would be advised to work more and publish in a Tier I journal. Naturally, after the student graduates, neither the student or faculty has the time or inclination to work on it. For many departments, small is beautiful. I think these attitudes have to be revised first before any perceivable change can occur in the overall output of IISc/IIT and catapult it to be in the top 100 internationally.

Monday, December 29, 2008

Publish in Nature and Wikipedia

Anyone submitting to a section of the journal RNA Biology will, in the future, be required to also submit a Wikipedia page that summarizes the work. I am surprised at Nature preferring Wikipedia over Scholarpedia

Thursday, December 18, 2008

Indian Science Academies

In an article in Current Science titled "Losing out on Fellowship of Academies", Professor K. R. Rao suggests that the success of being elected as a fellow of the academy depends primarily on how well one can network in the community. Professor CNR Rao comments, "There are many people in this country who look down upon people who publish papers or comment on their work. In fact, I am sorry to say, some of the Academies have encouraged the art of non-publishing"

Wednesday, December 17, 2008

Rejection Letter Genres

There is an interesting discussion on rejection letter genres. I collected a few of them for laughs. IISc normally follows the Black hole rejection model or the Don't Shoot the Messenger model (see below). When I had applied for a faculty position in IISc in late 1997, I got the job. The offer letter followed the Don't Shoot the Messenger model, " I am directed to inform you that you have been selected..." But when I had applied in late 1996, I got a rejection letter of a different genre that is listed below. I will leave it to you guess what it was.

  • It's Not You, It's Us - "You really truly are special and a unique and valued human being, but we really had an astounding array of highly qualified applicants and..."
  • Postcard or Mass Email - "Your application to Snooty U. is no longer active."
  • You All Suck - "We had a kazillion applicants. None were suitable. We closed the search."
  • Baby Announcement - "We don't want you. But let us tell you all about the Chosen Person..."
  • Weirdness - "We had a bajillion applicants. None were suitable. But you were great but sorry"
  • Dear John (Jane) Break-Up Letter: "It was great getting to know you better and let's stay friends..."
  • The Black Hole Rejection - apparently the sheer density of Important Academic Minds all located in such a small area prevents rejections ever from leaving the desk of the search committee chair.
  • Don't Shoot the Messenger - "The search committee has asked me to inform you that you are no longer under consideration."
  • The pleasure is ours: We want to thank you for giving us the opportunity to review your incredibly interesting materials that you put together to apply to our job, which we won't be considering you for anymore.
  • No Social Skills: "The SC has formulated its short list and I am sorry to report that you are not on it."
  • Ooops! There's no $ for this Position - We know we put you (and others) through phone interviews, campus interviews, checked all your references, but now it appears that we do not have the $ for this position after all.
  • HR Robot - "Dear NAME OF APPPLICANT, Position #24601Q55X has been filled. Please see [HR WEBSITE HERE] to apply for another position."
  • We're Totally Lying - "Dear DR. WRONG-NAME, After carefully reviewing your materials with the utmost consideration possible, we regret to inform you..."
  • Odd Euphemisms[We wish you a beautiful day and the best of fortune in your journey towards employment..."]
  • The Make-Us-Feel-Good-About-Rejecting-You "Please accept my sincere wishes in your job search, and I am sure there is a high probability that you have already signed a contract for another position elsewhere"
  • Keep Your Chin Up "We would also like to emphasize that our decisions had at least as much to do with the needs of our department as with our evaluations of an applicant's qualifications, and hope that our decision will not discourage you as you continue on in the job market process."
  • Can’t Write My Way out of a Paper Bag. Thank you for your interest in the faculty position at X. We have completed the search and have selected an individual for this position.
  • Tautological rejection letter. The Search Committee for this position met recently to review all the applications, which were received. Unfortunately your application was not among those, which were chosen for the short-list of candidates.
  • Social worker's rejection letter. Our decision should not be read as a judgment on the quality of your work or your abilities.
  • Too Toxic. Thanks for applying. Here are your application materials. You might need them in the future.
  • Diversification. We used you and a bunch of other people to apply but had no intention of hiring anyone since we had someone already on the faculty/staff (or their spouse/lover) who we needed to transfer and just needed to convince ourselves that we are committed to diversity.
  • Old Friends. Although you are a stellar candidate and we should be kissing the bottom of your shoes to get you here, we prefer to hire the ABD candidate with no pubs, grants, or presentations because he is an alumnus of our university.
  • That Which Shall Not Be Named. On behalf of my colleagues and myself, let me thank you again for your application and for the chance to look at your work so closely. I am sorry to have to bring disappointing news. What news?
  • Friends and Family. We are pleased to announce the hiring of Professor Great White Hope from Awesomest University, who has done X as well as the Y we really need and who is moreover an alumnus of our department, returning to us after a two-year stint at Awesomest. We look forward to adding him to our marvelous faculty. Good luck in your job search.
  • The Double Down, two rejections in one letter. "I am writing to let you know that we have successfully completed our search for the position in our department for which you applied. This process took considerably longer than we had anticipated because a second, unanticipated position came open in our department, and we also considered your application for this second opening. I want to thank you for giving us the opportunity to consider your candidacy twice.
  • Insincere Sympathy. "I know from personal experience how painful the job seeking process can be and wholeheartedly hope that you find a position in the future."
  • Do I need to open this? An email saying, "Please see the attached file regarding your application." The attached file was called, "MyNameRejection.doc"
  • Monday, December 15, 2008

    h-index

    The Hirsch index is probably the most interesting measure provided in scientometrics in recent years. Some statistics based on h-index for IISc

    Number of faculty who have h-index more than 80: 1 (Prof. CNR Rao)
    Number of faculty who have h-index between 50 and 79: 1 (Prof. P. Balaram)
    Number of faculty who have h-index between 40 and 49: 2
    Number of faculty who have h-index between 25 and 39: 26
    Number of faculty in engineering who have h-index more than 30: 0
    Number of faculty in engineering who have h-index more than 25: 3
    Some interesting correlations that seem to be valid in IISc. Np > h*5; Nc > 3*h*h, where Np and Nc refer to the total number of papers and citations.

    However, one area in which h-index fails is for a young scientist, who is starting his career. In fact, h-index below 5 has no significant meaning. Therefore, a new method has been proposed to modify the index for graduate students and young faculty.

    Impact factors and Citations

    In an article, "Should we ditch impact factors", the problems of using the impact factor of a journal in deciding the quality of a paper is discussed. A journal's impact factor is calculated based on the total number of citations of all its papers published during the previous two years, divided by the total number of papers.

    The logic behind using an impact factor of the journal in determining the quality of the paper is that we believe that journals with high impact factors publish high quality papers while lesser quality papers are published in a 'low impact' journal. This argument is fundamentally flawed because several analysis have shown the 20-80 principle to hold i.e., only 20% of papers published result in 80% of the citations in that journal for that year.

    The current way of determining the impact factor of a journal can be manipulated. As an editor of a polymer journal, I know ways that editors/publishers manipulate to reach a higher impact factor. For example, because the impact factor is calculated for two years from Jan-Dec, reviews (which typically receive more citations than other papers) are published in the January issue of the journal.

    My detailed analysis of publications from IISc showed that a paper from IISc in a journal received more citations than the average number of citations for all papers published in that journal. Therefore, citations is a much better parameter to be used than impact factor. If a paper receives 20 citations after being published in a journal of impact factor of 2, it should be considered better than the paper receiving 10 citations in a journal of impact factor higher than two. Of course, citations should be compared in the same area of research or suitably weighted.

    The corrected impact factor, of course, is given by this :-)

    Wednesday, November 26, 2008

    Gender Gap in academics

    Gangan Prathap, the vice-chancellor of CUSAT, writes about the gender gap in academics. He states that in terms of financial (net-worth) returns, ONLY in the very long term is it better to be a man. He concludes

    By working with a sample where men and women have nearly identical backgrounds, it is seen that between the ages of 30 and 40 years, women actually have a positive gender pay advantage and lose this only after 40 years of age.


    Despite this, I find that in IIT/IISc, though nearly 50% of the students in chemistry and biology are women, less than 10% of the faculty in these departments are women.

    Monday, November 24, 2008

    Increase in number of Ph.D's

    Subsequent to my previous post, a colleague of mine asked me about the number of chemical engineering doctorates produced by India as to whether it was unusually low in 2006. In the period 1996-2005 (ten years), IISc produced twelve Ph.D's while the rest of IITs produced around 300. To put it in another way, the number of chemical engineering doctorates produced in India during the ten year period is one-third the total number of chemical engineering doctorates awarded by USA in 2007. However, here, we believe that small is beautiful :->

    On a happier note, the number of doctorates awarded in sciences and engineering in USA has increased from 29,855 to 31,801, an increase of 6.5% from 2006 to 2007. Chemical engineering has shown a moderate increase from 891 to 921, while a couple of branches (chemistry and mechanical engineering) has seen small drops in the number of Ph.D's awarded. Surprisingly, one of the largest increases has been in electrical engineering. With the President-elect promising increased funding to the sciences, I will be glad if these numbers continue to increase.

    Tuesday, November 18, 2008

    Number of doctorates...

    My colleague, Abi, writes in his blog about the number of Indians studying in US universities. The online newspaper, Rediff, mentions that India continues to be the the leading country to send maximum number of students to the United States for the seventh consecutive year.

    Some interesting stats from the survey of earned doctorates and from articles by Gangan Prathap, Rangan Banerjee, Highered and a host of other places (all stats are for 2006)


    Number of Ph.Ds awarded in USA: 45,596 (20,539 female)
    Number of Ph.Ds awarded to Asians in USA: 11,706 (4,345 female)

    Number of Ph.Ds awarded in USA in engineering: 7,191
    Number of Ph.Ds awarded in China in engineering: 8,573
    Number of Ph.Ds awarded in USA in engineering to Indians: 1,200
    Number of Ph.Ds awarded in India in engineering: 925
    Number of
    Ph.Ds awarded in India in engineering by IIT/IISc: 434 (65 by IISc)

    Number of Ph.Ds awarded in USA in chemical engineering: 893
    Number of Ph.Ds awarded in USA in chemical engineering to Indians: 149 (guessing 893x1200/7191)
    Number of Ph.Ds awarded in chemical engineering by IIT/IISc: 34 (0 by IISc)

    The above shows that China has crossed USA in the total number of doctorates awarded in engineering. Contrast this to India, where the number of doctorates awarded to Indians in India is less than that of the number of doctorates awarded to Indians in USA.

    Saturday, November 15, 2008

    Economics and research

    In a recent article, Professor Balaram wonders about the (in)effectiveness of mathematics and physics in modeling financial markets. In an another article in NY Times, the authors wonders whether the financial market crash would lead to the best and brightest, who had gone into Wall Street, to do scientific research and maybe even solve global warming.

    In his blog, the Ponderer says

    The Wall street guys provide a valuable service - but that service remains, briefly summed up, they move large amounts of money around, and make the process a little more efficient and supposedly a little more safe (perhaps not?). In no way or form the benefit to society even remotely approaches those of teachers, or medical doctors or scientist working on eradicating deceases or creating a new device. Wall street guys get compensated with a lot of money because they are in charge of moving huge amounts of money.
    Compensation for employees on Wall Street averaged $399,360 in 2007, whereas an average professor in a top university in the US earned less than half of that.

    What does the economic crash mean for Indian research? Prior to the economic/IT boom, in the 1990s, nearly 80-90% of the graduating B.Tech students from IITs used to go abroad for higher studies. In 2007, this number has dropped to 16%. Even worse, the numbers who continue to pursue their core profession (mechanical, chemical, civil engineering etc.) has significantly dropped in the last few years. However, this economic fallout has resulted in the number of people writing GATE to increase by nearly 50% this year. Maybe more students may likely consider research and education to be possible career positions but unless the attitude that "You get to do what you love and thus money should not matter" does not change, it is unlikely to happen.

    Thursday, November 13, 2008

    The 20/80 principle...

    The Pareto principle (also known as the 80-20 rule) states that, for many events, 80% of the effects come from 20% of the causes. Pareto was the first to observe that 80% of income in Italy went to 20% of the population and found it to be valid for the economic distribution in various other countries. Pareto developed the concepts in the context of the distribution of income and wealth among the population. Actually, the Pareto principle was first conceptualized by Juran. His classic book, Managerial Breakthrough, was the first book to describe a step-by-step method for breakthrough improvement, which is the basis of Six Sigma today.


    A parameter that is related to the Pareto's principle is the Gini index. If 80% of the effect come from the top 20% of causes, then the Gini index is 80-20=60. If 70% comes from 30%, Gini index is 70-30=40 and so on. If 50% comes from 50%, then it is the ideal with a Gini index of 0. The Gini coefficient is primarily used to determine the level of inequality in the income levels of the country. India's Gini coefficient, which was around 30 in 1960's and remained constant around that value till 1990 has risen from 30 to nearly 42 in 2003. China has also increased its Gini coefficient from 30 from 1980 to 47 in 2003. USA, Mexico and Brazil have higher Gini coefficients than India. The lowest Gini coefficients can be found in Scandinavian countries and Eastern Europe countries. Higher coefficient is bad, if you are a socialist, and indicates disparity of wealth.


    How can these parameters be used in a research institution? Inspired by our director's article, I set out analyzing the trends in journal subscription. It was determined (see detailed analysis here) that the scientists in IISc publish 80% of the papers in 20% of the journals that IISc subscribes to.


    Can this be extended to determine the effectiveness of an institution? For example, Gangan Pratap, the vice-chancellor of CUSAT, proposed a technique based on h-indices to evaluate institutions [In case you are interested, h1=71 (for the period of 2000-2004) and h2=26 for IISc]. An additional parameter could be the Gini index, which shows the inequalities within the institution. An analysis indicated that 77% of the publications for the period of 2002-2007 came from 23% of the faculty in IISc, effectively resulting in a Gini index of 54. A similar analysis indicated that MIT had a Gini index of 32. Of course, Gini index will be useful only in determining inequality in good institutions i.e., with high h1 and h2. One will have a Gini index of 0 when no one publishes ! When comparing two institutions with similar h1 and h2, maybe the Gini index can provide information on the "equality" of all faculty.

    Wednesday, November 12, 2008

    Entrance Exams in India

    In a recent blog, my colleague worries about the tinkering of the cutoff marks of the JEE exam. JEE is the exam which secures you admission to the prestigious IIT for the undergraduate program. GATE is the exam which secures you admission to IIT/IISc for the graduate program. Roughly 3 lakhs and 2 lakhs write JEE and GATE, respectively and apply for around 4000 undergraduate and 2000 graduate seats, respectively. Thus, it is the top 1% in each exam who finally gets admitted in IIT/IISc.

    In 2006, the cutoff in each subject for JEE was based on the average minus one standard deviation. In 2007 and 2008, it was based on the 20 percentile such that one has to score in the top 80 per cent in each of the three subjects to be considered for admission. In JEE-2009, the first cut-off will not be at the 20th percentile, but at the average score in each subject. Considering the distribution of the marks in JEE is a Weibull distribution, we know that the average does not correspond to the 50 percentile. The cut-off formula is changed just based on public criticism. For example, when the cutoff was based the 20th percentile in 2008, they corresponded to just 5, 0 and 3 marks in math, physics and chemistry, respectively. This raised a hue and cry that someone who scores 0 in physics but enough in mathematics and chemistry such that the aggregate exceeds the overall cutoff would get admitted. Therefore, it was changed to the “average” marks in each subject. There is absolutely no consensus whether sectional marks should be or not be used for cutoff.

    This kind of confusion and uncertainty has significantly affected the sister exam, GATE. GATE moved from a descriptive + objective exam to purely an objective exam after a detailed analysis of the marks obtained by the candidates showed that the relationship between marks obtained in the objective section of the paper with the marks obtained in the descriptive section of the paper was linear. GATE score was introduced in 2005 to differentiate candidates who secure between average plus one standard deviation to average plus seven standard deviations. However, this was the most misunderstood parameter even among my colleagues. A student with a GATE score of 700 was thought to have scored more than twice that of a person who scored 300. Actually, GATE score of 700 corresponds to average + 5 standard deviations while 300 corresponds to average + 1 standard deviation. I am glad that GATE has decided to give the actual marks obtained by each candidate, in addition to the GATE score, percentile, All India rank, starting this year (2009).

    One thing that I have been pushing for is the introduction of sections in the GATE paper. For example, the chemical engineering paper is roughly divided into seven sections of roughly equal weightage. Math, Fluid Mechanics, Reaction Engineering, Heat and Mass Transfer, Process Control, Process Calculations and Process Economics/ Industry with near equal distribution of marks in each section. There is no faculty in IISc who does research in the last four topics. Therefore, a candidate who secures high marks in the last four sections and nothing in the first three may still qualify in the top 1%, get admitted and find no one doing research in his/her favorite field. This problem is more acute in Mechanical, Civil and some other branches of engineering where the distinctions between the sections are even more drastic. Why can not GATE give marks obtained by the candidate in each section? The particular department in IIT/IISc can then choose which section should be given more importance? It is already done in CAT/GRE where marks in each section are given and the admitting department/institute makes the choice. It is not done because of the problem encountered in JEE. Should each section have a cutoff? If so, should it be based on 20th percentile, average or the 80th percentile. Which candidate should be ranked higher? One who scores 10 marks in each of the seven sections or one who scores 15,15,15,15, 10,0,0? The former is well rounded while the latter is better suited for your research.

    Going back to JEE, the statistical questions filled by the candidate when he writes the JEE exam shows that 98% underwent coaching (I think the other two percent lied) and 80% are from an urban population. It is time that JEE is significantly modified to allow less privileged people (who can not afford the coaching) to qualify in the exam. I know that one of the measures that is being thought of is to only allow candidates who have represented the country in the (Math/Physics) Olympiad OR secured an average plus two standard deviations in their respective board exam to write JEE. Because most of the board exams follow a normal distribution, only the top 2.5% will
    be allowed to write JEE. JEE will not be an objective exam (i.e., multiple choice questions) but will include several descriptive questions with marks ranging from 1 to 10 for each question. Subsequent to this exam, JEE will announce the marks in each subject for every student that took the exam. The departments in each IIT will decide the weight to each subject. The candidate then decides which branch to apply and to the IIT that (s)he likes. Computer science may give 60% weight to marks in maths and 20% weight each to physics and chemistry. The maths department in IIT may give 90% weight to marks in maths and the candidate can join for the Integrated M.Sc (Maths) there.

    How does it help? There are two independent studies that show the performance of the student in the B.Tech directly correlates with the marks obtained in the board (12th) exam. Similarly, one study shows that there is a direct correlation between the performance in the B.Tech class to the performance in M.E/M.Tech in IISc. No such correlation exists with the JEE/GATE rank. The department of IIT/ IISc gets to choose the candidate they want depending on the weightage they assign to each section.

    However, until significant changes are made to the entrance exams, we should be content that both the exams are conducted fairly without any political or other influences.

    Monday, November 10, 2008

    Makes me wonder

    A recent report submitted to the Planning commission of India laments that the total number of Ph.D’s in engineering is very small in spite of “good” academic salaries. In fact, the report cites the study that discusses the faculty salaries across various countries. This study ranks India the highest in terms of the Ratio of Average Monthly Faculty Salaries, in World Bank Parity Dollars, to GDP Per Capita. This ratio for India is 8.73 compared to USA(ranked 12th), which is 1.67. Therefore, the report to the Planning commission wonders why do not more people enter academics. The report wonders why only 1% of the graduating B.Tech class of an IIT opts for a PhD in India.

    At this time, it may be interesting to dwell on the statistics. The sanctioned strength for the number of undergraduate (B.E/B.Tech) is 5.6 lakhs per year. Roughly, 4.6 lakhs actually graduate with engineering degrees every year. Last year, 2.4 lakhs among these were picked up by the IT industry, 1.95 lakh students write GATE, 24 thousand write GRE. The total number of Ph.Ds graduating in India is around 1000, with IITs/IISc graduating 50%. Thus, roughly only 0.2% of the engineering students actually do a Ph.D in India every year. Therefore, I am indeed surprised that as high as 1% of the graduating B.Tech class of an IIT actually opts for a Ph.D in India.

    Regarding the salaries, the authors of the report to the planning commission must instead compare the salary of a professor with ten years experience in IIT with the salary of a B.Tech from IIT. My guess is both would be drawing the same salary. Considering this, it is not a difficult choice for a B.Tech not to choose academics. This has nothing to do with GDP per capita.

    Having said that, I would guess that roughly 1% of the B.Tech graduates from IITs do become faculty in IIT or IISc after completing a Ph.D abroad. This is not because the salaries are good compared to the per capita GDP but because IIT/IISc provide you with a great research atmosphere to work.

    Tuesday, September 30, 2008

    Being organized

    One big difference between being a student and a faculty member is the number of things that needs to be done. Several of these are small items - like sending someone an urgent email/calling someone to schedule a meeting (especially if you are a member of several committees). In addition, there are deadlines for proposals, manuscripts, reviews etc. I write 20 manuscripts a year and review around 50 manuscripts (my reviews are always thorough and detailed, even if it is to accept a manuscript). Keeping track of all of these requires you to come up a good system that works without follow-up. I use the Gmail account that retrieves emails from my university account, and I use RTM (remember the milk) for reminders and events. While I do like google calendar, it does not work like RTM because some of these activities (like reviewing a paper, finishing a manuscript) do not usually have strict deadlines and can be postponed. With the new Firefox extension that incorporates RTM into GMail, I have shifted to gmail permanently. Whenever I get an email, I star it and it is automatically marked in RTM as an event that I will be reminded later. Because of the gmail filtering system and the RTM addon, all my emails are automatically marked for action as soon as I open my mailbox (except for emails without a subject, which are automatically trashed).