I hate the “I just need to polish the edges” asymptote
Image
Reply
Author Archives: ibartomeus
Linkage Rules in Plant-Pollinator Networks
I have a new paper in PLoS One with open text, data and code. I like this paper for several reasons. First, is the fruit of the 20% rule. That is, using 20% of my time to side risky projects. I decided to take a webinar on hierarchical models to estimate occupancy last summer. I am not sure why I did it (well, it was free), but the idea of incorporating detectability processes to study animal occupancy was appealing to me. However, I was pretty sure I will not use this models any time soon and I had quite a lot of other things to do, but I decided to “lose” a week anyway attending to it. I didn’t make the connection to apply this concept to plant-pollinator networks right away, but like a month later I got the “aja” inspiration, and I realized that those models can be applied to networks as well. And this resulted in my first single author paper. Not bad!
The basic idea is that doing field work is very time-consuming, and you can not watch all plants species for an infinite time, hence the chances that you detect all pollinators visiting your target plants are low, specially for rare pollinators. Hence, when trying to model what makes a pollinator decide to visit a plant or another, you can not use this raw data because rare species will appear as specialists, when they are probably not. But I show in the paper how to bypass that limitation. You will have to read it to know how, but it involves the above mentioned modelling approach.
I also like the paper for second reason, and is that I think I made a good point on seeing the network as dynamic process. The models use floral traits to predict visitation. Hence, you can use this models to predict which pollinators will visit an invasive plant entering the network. Moreover, I show that I can predict re-wiring of links after a species is removed from the network. Cascading extinction models are quite popular, but do not incorporate this dynamical effects. I know that Jeff Ollerton et al. (who I hope is reading this) have some experimental data that test this “re-wiring” after a plant extinction, so I would love to see if the models predict their data correctly. I think the next big thing will be to incorporate a more dynamic view to p-p networks.
In any case, If you are field biologist don’t be afraid for the “hierarchical modelling” part, (which is not bayesian, but uses likelihood!) and give it a read, because I think I managed to explain it in quite plain English.
signing reviews pays back (and about sharing good and bad news)
Quick post to share an awesome experience I had today. I received an email from an author I just reviewed a paper. The paper was rejected. To my surprise that was a “thank you” email. I feel I have to quote it, hope that is ok…
I write to thank you for all the comments and suggestions. They have been extremely helpful in improving the quality of the manuscript and in calling our attention to previously unnoticed weaknesses.
I have been signing reviews for a couple of years now. So far I had one “thank you” letter, and zero angry letters. If I didn´t convince you before, are you still not convinced to sign your reviews now?
On a side note I realize I tend to share the good news, but not always the bad ones. However we should do it too. Twitter and blogs also work a bit like an empathy box and is good to share new cool papers and experiences, but also is good to share rejections (Yes, for example last week Proc B rejected my paper without review) or experiment failing (The aphids that were supposed to be my herbivory treatment, were ate by coccinelids), specially to show PhD students that everyone has ups and downs, and struggles to do science. Now I have to go to try to fix the aphid issue …
Peer-Review, making the numbers
We know it, the system is saturated, but what are we doing? Here are some numbers from 4 recent Journals I reviewed and published (or tried to publish) recently.
| Time given to me to complete the review | Time to take a 1st decision in my ms | |
| PNAS | 10 days | > 3 months |
| PLoSOne | 10 days | 2 months |
| EcolLett | 20 days | 2 months |
| GCB | 30 days | 2 months |
I think most reviewers do handle the ms on time (or almost on time), and that editors handle ms’s as fast as possible, so where are we losing the time? On finding the reviewers! In my limited experience in J Ecol I have to invite 6-10 reviewers to get two to accept, and that imply at least 15 days delay at best. And note that all the above are leading journals, so I don’t want to know how much it take for a low-tier Journal.
However, the positive line is: There are people willing to review all this papers. Seriously, there is a lot of potential reviewers that like to read an interesting paper on their topic, specially if they get some reward other than being the first on knowing about that paper. So I see two problems, which rewards can we offer and how to find the people who is interested in reviewing that paper efficiently.
1) Rewards: Yes, I love reviewing, I learn and I feel engaged with the community, but it also takes a lot of time. However, a spoon full of sugar helps the medicine go down. I don’t want money, I want to feel appreciated. For example, Ecol Lett offers you a free subscription for 3 months or GCB a free color figure in your next accepted ms (given that you manage to get one accepted). I am sure other options are out there, including some fun rewards, like for example “the person with more reviews in a year wins a dinner at next ESA/Intecol meeting with the chief editor” to put a silly example. Recognition is another powerful reward, but more on that line in the next item.
2) Interests matching: Rather than a blind guess from the editor of who will be interested in reviewing a paper, we should be able maximize interests. Can we adapt an internet dating system for finding a suitable partner to find a suitable reviewer? As an editor, I would love to see which reviewers with “my interests” are “single” (i.e. available) at this moment. Why sing in as a reviewer? May be because you want the free subscription to Ecol Lett or you die for this dinner with Dr. X. Also, by making your profile and activity public is easy to track your reputation as a reviewer (and of course you can put your reputation-score in your CV). Identify cheaters in the system (which submit papers but don’t review) will be also easy, and new PhD students can enter the game faster. Any entrepreneur wants to develop it?
While there is still also a lot of bad advice out there which contribute to saturate the system, other models to de-saturate the system are possible (PubCreds are an other awesome idea). I am looking forward to see how all it evolves.
Native bees buffer the negative impact of climate warming on honey bee pollination
We have a new paper in GCB lead by Romina. In this paper we do a very cool thing. We characterize the daily activity period of a bunch of bee species and how this activity is modulated by temperature. We show that while honeybees decrease visitation to watermelon at very high temperatures (literature suggest that the reason is that honeybees need to go for water more often when hot, hence have less time to visit flowers), some native bees concentrate its visits on the warmer hours. I think that understanding behavioural differences among species is neat to answer BEF questions.
In addition, we play a bit with future temperature scenarios to see if (all else being equal) visitation and pollen deposition will change with warmer temperatures. We show that the visitation reduction predicted for honeybees is compensated by an increased visitation rate by native bees (taken altogether). Despite this predictions should be interpreted with care, it adds up to the several lines of evidence suggesting that conserving all species is needed in order to have flexible ecosystems able to cope with environmental change.
Ramon Y Cajal advice
This post has two purposes, first, celebrate that I was awarded a RyC fellowship to go back to Spain, which is very exciting. Second to recommend to everyone the reading of Ramón y Cajal advice for a young researcher [PDF here].
It was written in 1920’s and is surprisingly modern. He makes a strong argument to let the data talk for your science and he make some very relevant points against the inclusion of honorary authors. I also love his steps to write a paper:
(1) Have something to say, (2) say it, (3) stop once it is said, and (4) give the article a suitable title and order of presentation.
He is a little bit too harsh on substituting talent by working hard, but I agree that working hard (i.e. don’t expect discoveries to come easy) is a good advice. Putting that together with his advice on how to criticise others work without hurting any feeling (i.e, always acknowledging the good points first), I can summarise it with a quote borrowed from my father: “work hard and be nice to people”. On my own experience, I recommend anyone to maximize the feeling that Science is a big community of helpful people with a common purpose rather than a competition among researchers.
The advice for the Spaniards (how to do science from a country on the cue of scientific production and with very limited funding in 1920) is not as up-to-date nowadays, but I am affraid we will have to apply some of his advide on that soon, if things keep that way.
I don’t agree with everything. For example, I think working in group and establish collaborations is basic to get the most of our imagination and talent, instead of working alone for long hours. I also think is funny the advise he gives in order to find an appropriate wife, and it may look even a bit offensive nowadays, although the bottom line is quite true: find someone who understands you!
The last thing I want to highlight is that i love how he transmit the ideal of a scientist as a nobel pursuser of the truth, unbiased, humble, honorable, almost kind of a knight extracted from a tale. But I’ll let you read the rest. Enjoy.
TraitBank
In brief: Who is in to create an Open Trait Data repository?
In this same moment at least 10 researchers (but mostly undergrads) are compiling trait data for some exciting analysis. That includes myself. In fact, most trait analysis are hampered by the quality of the traits, which are often lumped to the species level, and hence do not capture the natural variation, or info for some species is based on just one population with the hope that it is representative. Paradoxically, I think this trait data is very abundant, but not available. Thousands of researchers have measures, for example, of body size for a bunch of specimens of his/her preferred taxa. This data is just not accessible or is scattered on the net.
There are some databases (some open, some not) with traits for some groups (plants, birds and mammals) but not a joint effort to capture all this knowledge like the GenBank initiative. So I propose to create a TraitBank. The technology is easy to implement (from a SQL server liked to a web, to a simple Google spreadsheet), but the key aspect would be to enroll the community to make trait data deposition encouraged upon manuscript acceptance. Do you think that the leading journals will ask authors to deposit any morphological or life history measurement reported in the paper? It will also be important that a well-known independent organisation host the data. Any idea on who to contact? would Figshare be an option?
The fields should be very delimited to allow an easy search and compilation of information; as a first pass I would propose:
– Publication associated with the data and/or author
– Species taxonomy (full taxonomy can be retrieved from ITIS)
– Measurement is in wild or captive populations
– Region and Lat/Long of the measurement
– Category (morphological;life history; or ecological trait)
– Subcategory (e.g. body mass; clutch size; survival; phenology…)
– Mean value, SE and n: Units should be fixed by the subcategory.
A form and an option to upload a large csv should be enough. An API that allow connecting to R would be a blast. So how can we move that idea forward?
Food webs: reconciling the structure and function of biodiversity. Really?
I read this paper (Food webs: reconciling the structure and function of biodiversity; Thomson et al. 2012 Trends in Ecology & Evolution, 27(12):689-697) with great interest because the title is really promising. Indeed it is nice overview of what’s out there in terms of network and functioning, but not much reconciliation. First I have the feeling that community ecologists (even if they don’t use network metrics) are already (and have been for a long time) on the framework they describe in Table 1C. But my main concern is that I missed an answer to the question: What can a network approach add to the study of ecosystem functioning?
Well, I have two ideas that can help answering that.
1) Network approach can be very useful when the function itself is defined by the network. If you are studying pollination or pest control, the actual function delivered is contained in the network structure, hence species richness, diversity or composition (or new metrics, like FD) can be unable to fully explain functionality because what confers high levels of function (or stability) to the community is the network properties (e.g if it’s modular, generalized or well connected). I know some pople is on that path, so I am looking forward to see what they find.
2) Another situation where networks can make the difference is when indirect interactions modulate the function, but are too complex to track them one by one. Networks can describe better phenomenons like apparent competition or cascading effects than any other classical approach. If this type of complex interactions are relevant for the level of functioning measured (e.g. productivity of the basal level), then, adding the network perspective can be more informative than classical approaches.
May be what I am saying is too obvious, so the authors didn’t cover it, or I may be missing something, but this is the direction I would like to see things moving.
Happy biRthday
Today is my birthday. It’s also the birthday of a close friend. What an incredible coincidence! Or wait, may be is just expected. One more time R comes into our help, because it has a built-in function to answer our question.
Which is the probability of two coincident anniversaries among a group of 17 people? (yes we have a mailing list, so I can count my friends semi-objectively without the fear of not counting them all). Just type:
pbirthday(n= 17, classes = 365, coincident = 2)
The answer is approximately 0.3, that is 3 of every 10 friend groups (of that size) have at least two anniversaries that coincide. Not that impressive, isn’t it?. But the beauty of stats is that stats are here to correct your intuition. To have an impressive coincidence (and statistical significant) you will need a group of 47 people, none of them with coinciding birthdays. And then, probably nobody will be amazed.
qbirthday(prob = 0.95, classes = 365, coincident = 2)
Anyway, happy birthday to all readers celebrating today (if any)!
How I deal with work overload (without working too much)
I have several posts I would like to do, but this month has been very hectic. This encouraged me to revise how I deal with my tendency to overcommit myself with new projects, while managing to accomplish deadlines and still not working on the weekends. The result is that I am posting this instead of other alternative posts. See why.
1) Externalize fun work for out of the office. When I arrive home I play with my daughter so there is no way I can do actual computer work there. However most of my work involves thinking. I can think in many places. My favourites are on my bike on the way to work or when running. I not only think about work then, I also wonder about other stuff or picture myself in a tour de france time trial. However, I find that some fun problems are better solved in that context. Why? Because if you don’t come up with an idea in 5 minutes while sitting in front of your computer you feel desolated, but is ok to not have ideas if you are already doing something (e.g. running). Also, because if you are on your computer you tend to try (and put hands on) the first thing your intuition tells you will work. This way is easy to get lost in the details or do overcomplicated things that won’t work at the end. However, while running, you are forced to develop all the steps necessary and abstractly think if they will work discarding bad ideas way faster. Plus, blod is pumped to your brain continuously boosting your potential (or this is what I hope). But take notes as soon as you get out of the shower!
2) Minimum effort rule. I usually start by the task that requires less time to be completed. That way I can take it off my list and maximize the chances to move on any project. If I can solve something (e.g. a review or a simulation for a coauthor) in 1 to 4 hours, I just do it and have it done fast. Answering a question by email? I’ll do it asap and archive it. This short tasks are usually related to collborations and that also make happy those people and allow the project to keep moving.
3) Block time. The minimum effort rule fails when you start spending most time completing short tasks so there is no time left to work on long daunting (but exciting) projects. Then, I decided to block at least 2-3 full mornings or days a week to work on that kind of long term projects. No answering emails, no improvised meetings, no multitasking on the blocked time slots.
I thought about that post on my bike ride this morning, I knew it will be written fast, so I did it as soon as I have some spare time, but not this morning. This morning was blocked to do some other analysis.
I’m still on the process, so comment on what works for you!
Bartomeus lab 