230 likes | 243 Vues
Research practice. Define research topic Prepare research plan Conducting the research. Research practice. Define research topic Prepare research plan Conducting the research. Research practice. Define research topic Often part of a larger scheme State specific , testable hypotheses
E N D
Research practice • Define research topic • Prepare research plan • Conducting the research
Research practice • Define research topic • Prepare research plan • Conducting the research
Research practice • Define research topic • Often part of a larger scheme • State specific, testable hypotheses • an exercise in logic
Research practice • Prepare research plan • Updated review of the literature (existing grant proposal may be outdated) • an academic exercise • Types of sources • peer reviewed journals • non-peer reviewed journals • published reports • internet
Request to review To: shafir@agri.huji.ac.il Subject: reviewing BES-04-0168? Message-ID: <20040506093409.MCnarnia6469299@gemini.scholarone.com> From: giurfa@cict.fr Cc: Date: Thu, 6 May 2004 09:34:09 -0400 (EDT) Dear Sharoni Would you be interested in reviewing for Behavioral Ecology and Sociobiology the following manuscript: … If you would like to read the Abstract of this manuscript before committing yourself, you will find it pasted in further below. Please let me know whether or not you are willing to review this manuscript, by replying to my above email address. If you agree, I'll email you how to access the full manuscript. Looking forward to hearing from you soon, With best wishes, Martin Giurfa Associate Editor for BES
Thank you! From: giurfa@cict.fr Subject: Re: Review for BES-04-0332 Date: October 12, 2004 2:39:43 PM IST To: shafir@agri.huji.ac.il Thanks Sharoni! Martin From: sksakal@ilstu.edu Subject: RE: Review for Ethology ms ETH-05-0075 Date: October 24, 2005 4:34:13 PM IST To: shafir@agri.huji.ac.il Great! Thanks Sharoni - details to follow shortly. Scott Sakaluk Professor Phone: 309-438-2161 Department of Biological Sciences Fax: 309-438-3722 Illinois State University email: sksakal@ilstu.edu Normal, IL 61790-4120 http://www.bio.ilstu.edu/sakaluk From: courant@avignon.inra.fr Subject: AR:comments on manuscript (E1441) Date: April 15, 2004 10:09:38 AM IDT To: shafir@agri.huji.ac.il Dear Sharoni, Thanks a lot for your fast review of that manuscript. Best regards, Christiane --------------------------------------------------------------------- Christiane Courant Managing editor of Apidologie UMR INRA-UAPV "Ecologie des Invertébrés" Site Agroparc F-84914 Avignon cedex 9 Christiane.Courant@avignon.inra.fr Tél: +33 (0)4 32 72 26 19 ; Fax: +33 (0)4 32 72 26 02 Apidologie on line: http://www.edpsciences.org/apido
Submission From: Editor@abs.animalbehavior.org Subject: Your Submission to Animal Behaviour [7977] Date: January 30, 2004 11:32:49 AM IST To: shafir@agri.huji.ac.il Dear Dr. Shafir Caste-specific differences in risk-sensitivity in honey bees (Apis mellifera) Thank you for submitting your manuscript to Animal Behaviour. We have given it the number 7977 which should be quoted in all correspondence. The manuscript will now be assigned to an Editor and reviewed by at least two reviewers. We shall be in contact again when we have reached a decision. This usually takes about 8 weeks but please bear in mind that this may be longer during conference and holiday periods. Best Regards From the Editors of Animal Behaviour
From: Editor@abs.animalbehavior.org Subject: Your Submission to Animal Behaviour [7977] Date: March 19, 2004 10:58:06 AM IST To: shafir@agri.huji.ac.il Dear Dr. Shafir 7977 Caste-specific differences in risk-sensitivity in honey bees (Apis mellifera) I am happy to inform you that your manuscript is acceptable for publication in Animal Behaviour, subject to suitable revision in line with the comments given below. The reviewers raise important issues, but I would like you to pay particular attention to reviewer 1's point suggesting that it is variation-sensitivity, rather than risk-sensitivity, that your study measures, and also that a stepwise procedure should be avoided in the regression analyses that you perform. I also agree with reviewer 1 that the inclusion of the process-based explanantion using Weber's law does not really add anything to the argument presented, and that removing this would help to shorten a rather lengthy paper. However, if you feel this information is essential, perhaps you could provide more justification for doing so, so that its relevance is made clear. Please also include in the Methods section the ethical information provided to Dr Turner by email. When you revise your paper, you should prepare a detailed explanation of how you have dealt with all of the reviewers' and Editor's comments. Refer to the Instructions for Authors (on the main menu of Editorial Manager at http://anbeh.edmgr.com and in the back of the latest issue of Animal Behaviour) for details of our house style and for a list of file types that are acceptable for revised papers. Log in to Editorial Manager as an Author to submit your response to the comments and your revised paper. Changes in the revised paper should be highlighted; you can use Track Changes, then Highlight Changes in Word or just underline your changes. The revised paper will be edited and further changes may be necessary then. Final acceptance for publication depends on the date when the manuscript has been edited and all queries have been resolved. We should like to receive the revised paper within 120 days. If you think you will be unable to revise your manuscript in that time please let the appropriate Editorial Office know (justine.fox@nottingham.ac.uk for UK papers). Please do not reply directly to this email. Yours sincerely Louise Barrett Editor Response
Reviewer #1: This well-written manuscript continues the use of a promising experimental paradigm (proboscis extension response) to explore a novel question: do castes differ in risk-sensitivity? The integration both functional and mechanistic approaches is commendable. The experiments seem carefully conceived and well executed. The results are of general interest, and fit nicely with predictions. I two major quibbles with the work, some minor quibbles, and a few editorial questions, as follows: Major 1. It's not risk-sensitivity. What the authors claim to be risk-sensitivity strikes me instead as "variation sensitivity". Unlike the approach used by Shafir et al. (1999), where bees experienced a random sequence of rewards each time they selected the variable option, these bees experienced deterministic sequences (description begins on line 231). That is, they got "1-0-0-1-0-0-1-0-0-1-0-0-...." (for the low variance) or "1-0-0-0-1-0-0-0-1-0-0-0-1-....." (for the high variance). This is a problem, because the deterministic sequences ensure that there is no risk, only variation. While I agree that the authors may have shown a response to these treatments, it cannot be called "risk-sensitivity". No research in risk-sensitivity that I'm aware of used such strongly deterministic sequences to simulate a "risky" state. There's a big difference between 0-1-0-1-0-1-0.... and a random string of equi-probable 0's & 1's in terms of how organisms might perceive risk, and in the functional consequences of risky choices. A forager feeding at what Shafir et al. call low or high risk would experience IDENTICAL foraging gains after some period of foraging. There's no shortfall-minimization reason to be risk-sensitive, and there's no risk. 2. Stepwise models are flawed. All of the logistic regressions presented here are stepwise models. There is a lot of theory out there about why stepwise models are to be avoided when one is making inferences about variables (as opposed to finding an optimal subset of variables for the purpose of prediction, which is not the point of this research). Some of it's reviewed at: www.gseis.ucla.edu/courses/ed230bc1/notes4/swprobs.html. It strikes me that if the authors wish to test hypotheses about the importance (or non-importance) of different variables, then all of them should be in the "final" model. That is, a priori decide what variables to examine/control, then run the model with all of them in. Don't build an potential inferential "house of cards". It's true that each variable in the logistic multiple regression is assessed for its contribution to maximizing the likelihood of the observed data by comparing the model fit with and without it in the model. This sounds like a stepwise procedure, but it isn't. The other variables in the model are statistically controlled when the focal variable is assessed for its contribution to model fit. The choice of stepwise models makes me doubt the overall conclusions. That said, I suspect that the conclusions of Tables 3-5 and Figures 1-4 would not be seriously altered if the analysis is performed properly.
Minor In a few places (lines 310-311, 328-331) non-significant results are presented incorrectly. Bees did NOT show faster learning of the high reward relative to the low reward. Mean preference for the high reward was NOT greater in the 0-.4 condition than the .4-1.2 condition. I don't see the value of using precious journal space to present trial-level data, as done in Figs 1 to 4. This is particularly true when there aren't any obvious temporal trends of interest. Instead, any significant trend (Fig. 4) can be described in the appropriate Table. Tables 3 through 5 can also be made more informative (and not just by presenting full model fits, rather than stepwise-selected subsets of variables). In particular, the direction (sign) of each significant effect should be provided, along with an indication of what state (e.g., 0 or 1) was associated with what preference (e.g., low or high variance). This way, one could support previously unsupported statements like (line 318):"When bees extended their proboscis, they showed significant preferences for the high rewards, with a significant difference between conditions". lines 359-364. It's unclear to me whether this analysis (including Figure 6) is worth presenting. Why do we need to know that hungry bees ate more than less hungry ones? I'm not convinced that the "process-based explanation" using Weber's Law adds much to the study (other than 3 manuscript pages of discussion and a figure). This may simply reflect my "functional" bias (and "process" ignorance), but if there is a compelling reason for this lengthy discussion, it's not obvious to me. Note, however, that the other mechanistic approach (CV) strikes me as clearly relevant. Editorial line 86: remove extra bracket before "Rescorla". Methods. No information about number of honey bee colonies is provided. The authors note how many bees were used, but these are social animals, so number of colonies is also critical. line 131. What proportion of bees didn't extend their proboscis? Answers the nagging question: is this a tiny subset of bees that was tested? line 137. What proportion of bees didn't extend their proboscis? lines 214, 216, 219, 220. Add the word "discrimination" before "experiment", or it gets very confusing. line 269. US not previously defined. line 308. the meaning of "generalization" in this context is unclear. line 316. No statistical test provided to support the hypothesis of random choice.
Research practice • Prepare research plan • Updated review of the (peer-reviewed) literature • Databases • ejournals • Types of sources • peer reviewed journals • Impact factor? Rank? • non-peer reviewed journals • published reports • internet
Publishing “All the thinking, all the textual analysis, all the experiments and the data-gathering aren’t anything until we write them up.” “In the world of scholarship we are what we write.” “It is a truism that an experiment is not done until it is published.” Donald Kennedy Academic Duty
New and recommended
Research practice • Prepare research plan • Updated review of the (peer-reviewed) literature • Reformulate specific hypotheses • Design an experiment to address each specific hypothesis
Experimental design “Nothing is as simple as it may first appear ...”
Experimental design • Decisions, decisions, decisions ... • Type of data to collect • Size of sampling unit • Number of replications • Statistical analyses
Experimental design “Spend a lot of time planning -- it will save you a lot of time.”
Experimental design Ergonomics
Experimental design Automation
Experimental design “Nothing is random” • Haphazard • Random • Pseudorandom • tests: means and runs
Experimental design Real-world vs simulation • Real-world • smaller sample size • “less random” • Simulation • larger sample size (but with limits!) • “more random” (but with limits!)
Experimental design “Plan everything in advance -- during the experiment don’t think” • Micro • write precise experimental protocol • prepare charts of protocol • Macro • prepare chart of full experiment
Experimental design Constraints • Costs • Labor • Equipment • Time • Seasons • Other ...