All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Both reviewers agree that the manuscript is now suitable for publication and that it is a valuable introduction of improved statistics to behavioral studies. Please check the missing references mentioned by Reviewer 1. Although some small grammatical errors remain, the writing is now much clearer, and I believe that it will be comprehensible to readers.
[# PeerJ Staff Note - this decision was reviewed and approved by James Reimer, a PeerJ Section Editor covering this Section #]
The wording in the revised manuscript is more concise and accurate than in the first version, and incorporates the suggestions of the editor and other reviewers. For unknown reasons, two references were left off in the text: page 24: Hazlett 1981 (shells used by hermit crabs); Dembowska 1962; page 25.
Very novel experimental design, supplementary videos clear.
The authors have incorporated suggestions regarding their statistical analysis as recommended by another reviewer.
A novel approach to crab behavior. Most papers on the subject are strictly observational. The sharp videos are informative and far better than existing film clips that are available on line.
The authors have satisfied my comments. I have no further concerns.
Fine
Fine
I apologise for the time it has taken to complete my decision on your revised manuscript. I was busy and away from my desk during the holiday period and several manuscript revisions arrived at the same time, creating a backlog of work.
Both of the original reviewers were available to examine the revised version of your manuscript, and both considered it to be greatly improved. Reviewer 1 has only one comment for you to consider. Reviewer 2 has several suggestions to clarify the manuscript. (However, I do not agree that you need to shorten the Abstrat.)
Unfortunately, there are still a large number of errors in grammar, word choice and organization of material. These errors are far too many for me to correct. As an example, I have corrected the paragraph L121-145 (pdf attached). Every line has multiple problems. These problems would prevent readers from understanding and appreciating your interesting research and must be corrected before publication. Whether you hire a different editing service or find an English speaking colleague who would be willing to undertake this effort, is essential that the corrector takes the time for the large amount of work needed and is willing to correspond with you to be sure that the changes capture your intended meaning.
To facilitate future editing, please submit the revised version in 1.5 or double spacing, if possible. Single spacing makes it difficult more the editor to insert comments and corrections. Also, please remove the right justification, as indicated in the Instructions to Authors. Right justification results in variable spaces between words and make editing more difficult. The colored background for the Abstract is also not needed.
Because there are so many errors, I have not read the manuscript in detail. I will have to read it again carefully after the corrections, so there may be other corrections. Below are a few additional comments.
Specific comments
You expressed uncertainty about what I meant when I requested that you discuss the ‘strength of the evidence’. You are correct, of course, that evidence for the strength of the data is the statistical analysis. What I am requesting is that you also state in words whether the evidence for each finding is weak, moderately strong or very strong and point out specifically what the statistical evidence is and where (for example, which table) the reader will find confirmation of that evidence. Furthermore, it can be useful to discuss what some of the factors might be when evidence is weaker, for example small sample sizes for some size classes, inconsistent behavior of some individuals, or other aspects of the design or results.
I don’t think you define WAIC the first time you used the abbreviation in the text (L70). Defining it only in the Abstract is not sufficient.
Some errors in the references remain:
• In the three Akaike references, along with the original date of publication, you should give the original details about the journal, volume, etc. If you read them in an edited volume, then you also need to provide full details such as the editors and city of publication.
• Other edited volumes also lack information about editors and publishers.
• Capital letters are missing from some article titles.
The table should be inserted after the references and on a separate page. Put the heading above rather than below the table.
The figure captions should come after the table, followed by the figures. If possible, also include the captions with each figure (on the same page or preceding page). It is very hard for the editor to evaluate the clarity of the captions and figures if he has to keep switching between pages, especially if these pages are not adjacent.
New version is much more concise and to the point. References are good, figures and tables clear.
Reasonable. It would be almost impossible to conduct such a study in the wild.
Reasonable. Size-related differences in hiding behavior or camouflage occur in other families of crabs.
I note that the biggest crabs did not carry caps. Did the authors see any morphological change in the gripping dactyls of the pereopods or the chelae that might affect their ability to create a cap? In some majoid crabs, the largest males lose their hooked setae and also have heavy cheliped with a gape between the fingers, while smaller crabs have the setae and retain forceps-like chelipeds. Do these crabs have a terminal molt? If so, that might explain the lack of a covering--if the crab dies after mating or releasing eggs, there is no advantage to expending energy on camouflage.
Current manuscript is an improvement over first version.
See General Comments
See General Comments
See General Comments
Review of Revision
Individual behavioral type captured by a Bayesian model comparison in cap making by sponge crab
Comments:
=========
The authors have done a lot of work to meet the concerns of the examiners. The new title is much more appropriate. However some concerns remain for me, and some new concerns have been created. It should be relatively easy to address these concerns.
1) The addition of an explict formulation of conditional vs marginal likelihood is great, but would benefit from being better structured. I found it confusing and doubt that the intended readership would be able to dis-entangle it. For example:
a) Line 172. There is potential confusion here since the data are said to be iid, yet are then assigned a conditional (upon w) distribution for each individual, and hence can not be iid (at the conditional level).
b) Lines 172-184. This presentation would be clearer to me if the w latent parameters were given a subscript i. Note that equation (3) only shows a single w value, yet has an index on the x variables. I would recommend using p(x_i|w_i) in line 172.
c) Note that it is at the marginal level, equation (1), that the data are iid.
d) Line 177. I think it may be the case that WAIC is estimating the expected log-predictive density, not predictive density.
2) The authors continue to confuse me with their perspectives of Maximum Likelihood and Bayesian approaches, and replied to one of my concerns by writing:
"We are thinking about the maximum likelihood method from a Bayesian point of view, so used the sentence."
Maximum likelihood and Bayesian are two different philosophies! Possibly they view ML as a Bayesian approximation, which seems to be the viewpoint used in Watanabe's book.
However, this failing to distinguish between two fundamental statistical philosophies leads to some very misplaced concepts being presented, for example the justification of using WAIC rather than AIC or Rsq (lines 304-324). At the end of the day, the approach herein is Bayesian and so AIC and Rsq are not relevant since they are Maximum Likelihood measures of model prediction and fit, respectively. I couldn't see any request from the other reviewers for the additional discussion about AIC and Rsq. They can't be used here, so why discuss them? [Aside: Using ML, AIC can be used if it is the marginalized likelihood that is maximized. Rsq is not appropriate for hierarchical models, doubly so if the response is not normal.] See also the mention of Rsq in lines 338-342. Elsewhere?
3) While the formula for conditional likelihood is given, there does not appear to be any presentation or discussion of how the marginal likelihood was calculated. For example, in the Choice of Material Size, how is the marginal likelihood, p_{model}(x|w_0), actually calculated? That is, I'm presuming (?) that the conditional model is specified in Stan, so how was this marginal likelihood calculated. Since use of this marginal likelihood is a focus of this work, it would seem essential that the reader is informed how to evaluate it.
4) The Abstract has doubled in size and would benefit from shortening.
5) Strangely, despite its increased length, the Abstract now fails to mention the need to use marginalized WAIC. This is surprising, since the authors replied to one of my concerns about the orginal manuscript by writing:
"We think the novelty is in the model comparison approach using marginal-WAIC for the repeatedly measured behavioral data."
6) Line 336. Change "However, it is used with a software without the consideration on this point" to something like "However, it is generally more convenient to use conditional-level likelihood within most Bayesian software, with the unfortunate consequence that conditional-level WAIC is often used".
Overview
This study used a novel statistical approach, marginal level Widely Applicable Information Criterion, to examine whether there are individual differences in the behavior that sponge crabs use to attach sponges to their carapaces. WAIC allowed the authors to compare models that did and did not account for individual differences. Giving crabs a choice of artificial sponges of three sizes, they tested 30 crabs once and 8 crabs five times as well as video recording the behavior of 2 crabs five times. Crabs picked up sponges, sometimes removed pieces of then, created a cavity to fit their carapaces, and then held the modified pieces over their carapaces. Crabs made non-random choices of sponge, avoiding the smallest sponges; larger crabs preferred larger sponges and the largest crabs did not use sponges. Most of the crabs that removed part of the sponge were those that had chosen larger sponges, and larger crabs removed less of the sponge. Of those crabs that excavated cavities in the sponges, larger crabs made larger cavities. There was no relationship between crab size and time to complete the sponge cap. The hierarchical model performed better than the non-hierarchical model though the magnitude of the difference varied between measures.
We were fortunate to have one reviewer with expertise in the behavior of related crustaceans and one with expertise in the statistical approach. Both agreed that the study is well worth publishing and suggested a number of important changes. I have many more suggestions to be sure that the paper’s scientific contribution can be understood and appreciated by a broader audience.
Please note that I am not a statistical expert, so some of my comments may be based on misunderstanding. However, I think that my ability to understand the issues would be similar to that of many potential readers, so additional explanation may be required even where I have not understood the issue well.
Major Issues
1) Linking goals to background/knowledge gap, results, and discussion and abstract.
This study makes two broad contributions. First, it provides a statistical approach to identifying individual differences in behavior. Second, it provides information on sponge size choice, trimming, and excavating in relation to individual differences body size and potentially other individual characteristics. However, these contributions are not consistently presented, making the overall contribution of the paper ambiguous. For example, the title and abstract refer only to the statistical approach, but the discussion focusses only on the behavior.
• Revise the Title and Abstract to more completely reflect the contribution of the manuscript.
• Revise the Introduction to present a more balanced background to both contributions. If you decide that the primary objective is the presentation of a statistical method for identifying individual differences in behavior, you should start with a background on individual differences and what problems exist in previous approaches, then present questions about sponge crab behavior as the system in which you will examine individual differences. If you decide that the primary objective is behavioral decision-making in sponge cap acquisition, start with a background to sponge crab behavior and the knowledge gaps in that system, then develop your statistical approach to the problem.
• The Discussion should follow the same order of topics as the Abstract and Introduction. It must systematically state what conclusions you can draw from the study, discuss the strength of the evidence, and then place it in the context of previous literature.
• As noted by Reviewer 1, much of the information currently included in the Introduction and Discussion are not directly relevant to the study and needs to be deleted. Your study does not directly address the adaptive significance of the observed behavior. This should be reduced to a few sentences speculating on possible adaptive significance in the Discussion.
2) Individual differences in behavior.
I am not familiar with the use of the word ‘individuality’ to describe what you are studying. I believe that you mean ‘individual differences’ or ‘repeatability’. You do not provide sufficient background to the study of individual differences or the statistical approaches currently in use. This is a very active area of research due especially to the interest in animal personality. You should examine this literature more completely to see what statistical approaches have been used by previous researchers and the strengths and weaknesses of these approaches. I did a brief search on the Web of Science using the terms: “individual difference*” and statistic and animal. I found numerous relevant articles, a few of which I have copied below. Checking the key words and citations of these articles will help you find other, possibly more relevant articles. These articles will also help you use terms and key words in your articles so that your contribution will be easily found by other researchers.
• On the usage of single measurements in behavioural ecology research on individual differences By:Niemela, PT (Niemelae, Petri T.); Dingemanse, NJ (Dingemanse, Niels J.) ANIMAL BEHAVIOUR Volume: 145 Pages: 99-105 DOI: 10.1016/j.anbehav.2018.09.012 Published: NOV 2018
• Evidence for risk-taking behavioural types and potential effects on resource acquisition in a parasitoid wasp By:Gomes, E (Gomes, Elisa); Desouhant, E (Desouhant, Emmanuel); Amat, I (Amat, Isabelle) ANIMAL BEHAVIOUR Volume: 154 Pages: 17-28 DOI: 10.1016/j.anbehav.2019.06.002 Published: AUG 2019
• Individual and Sexual Differences in Time to Habituate to Food-stimuli Presentation of Potential Prey in Hyla japonica By:Tanabe, S (Tanabe, Shintaro); Kasuya, E (Kasuya, Eiiti); Miyatake, T (Miyatake, Takahisa) CURRENT HERPETOLOGY Volume: 38 Issue: 1 Pages: 14-22 DOI: 10.5358/hsj.38.14 Published: FEB 2019
• Roll with the fear: environment and state dependence of pill bug (Armadillidium vulgare) personalities By:Horvath, G (Horvath, Gergely); Garamszegi, LZ (Zsolt Garamszegi, Laszlo); Bereczki, J (Bereczki, Judit)Urszan, TJ (Urszan, Tamas Janos); Balazs, G (Balazs, Gergely); Herczeg, G (Herczeg, Gabor)SCIENCE OF NATURE Volume: 106 Issue: 3-4 Article Number: 7 DOI: 10.1007/s00114-019-1602-4 Published: APR 2019
3) Clarity of results
It is hard for a reader to fully understand what has actually been found. Sometimes, this is due to missing information and sometimes to the information being divided between the figure captions and the results text.
• The figure captions should fully describe the subject and contents of the figure, but the actual patterns of the data should be in the manuscript text. There should not be overlap between statements in the figures and text.
• Be clear about sample sizes. Sometimes, you mention sample sizes but it is not always clear how many individuals and how many repeated trials provided data for each section. It might be clearer to provide this information as a small table in the Methods or Results.
• For each section, you should address the strength of the evidence for patterns. For example, in sponge choice, how strong is the evidence for an effect of carapace width, lack of legs and for individual differences? Are the individual differences fully explained by carapace width?
4) Incomplete Discussion
Many of the findings are not fully discussed and some are not discussed at all, while there is more extensive discussion of topics that your study did not address. For example, L280-294 needs more information and better organization. Start with the evidence you have for an effect of size on sponge size selection. Objectively discuss the strengths and weaknesses of the evidence. Consider the variation within and among individuals, not just the statistical trend. Then relate your findings to the any literature on size selection of sponge caps or similar phenomena. Where your study provides unique evidence for which there is no previous literature, make the original contribution of your study clear. Where the evidence from previous studies is weak, for example due to the lack of appropriate statistical analysis, this can be pointed out. Finally, you can present some functional explanations for why the observed choices occur. It might be interesting to also speculate briefly on mechanistic explanations: how do the crabs know how big they sponge is? how do they decide how much to remove?
On L296-318, I do not see the relevance of a review of tool-making and learning in invertebrates. You could perhaps briefly make a case for cap production as tool-making with a couple of pertinent references, but a review is not appropriate here because this is not the focus of your study. On the other hand, you have not discussed at all your findings on cutting, cavity making, and latency to produce a cap. A good discussion would follow the type of organization I described in more detail for sponge choice above. Do you have evidence that differences between individuals are explained completely by size differences or are there other differences? How strong is the evidence? I would expect to find a discussion of individual differences linking your findings to the broader literature, not just cap-making crabs, because that was a focus of your Introduction. Finally, I would expect a substantial discussion of statistical methodology for finding individual differences. How successful was your method? Are other advances needed? To what extent have similar or alternative methods been used and were they more or less successful?
5) Clarity of presentation: English grammar
The use of English needs to be improved. I note that you acknowledge help from an editing service. However, there are a very large number of grammatical errors, so many that it is not reasonable for the academic editor to correct them all. Some of the most frequent include:
• Misuse of the definite article (‘the’) when an indefinite article (‘a’, ‘an’) or no article is needed.
• Inconsistent use of tenses, sometimes even in the same sentence.
I have attached an annotated version of the pdf in which I used highlights to indicate some of the errors that I noticed in my reading. However, there are probably more.
Other concerns
L20. ‘Remarkably’ is too vague a word; be more precise about the magnitude and uniqueness of the effect.
L27. Explain more completely how changes in behavior with size are related to individual differences. One could study size effects based on a single observation from each individual. Are you interested in individual differences that are independent of size?
L46. Is ‘sponge crab’ the same as dromiid crab? Please be clear what you are referring to when you introduce the term. I think the English family name is dromiid rather than dromid. Also L72.
L46-49. In this section, briefly describe the behavioral all the steps in obtaining a cap (selecting a sponge, cutting, excavating a cavity, holding the cap on the back. This will prepare the reader so that it will be clearer what you are referring to in the Methods (L131-133).
L53. ‘Data’ is plural.
L72. It is not clear how the McLay reference supports the topic sentence referring to several studies on individual differences in material preference.
L83. What is already known about the use of sponges and behavior of this species?
L83-96. The overview of your approach (L83-89, 92) is helpful for understanding the Methods section. However, you should not provide your results here (89-92). The statement regarding the effect of including carapace width as a variable (92-95) is important, but needs to be much clearer and should not include the results.
L92-96. It is important that in the Discussion you discuss more completely the evidence that the better fit of the hierarchical models is or is not due to size differences, not just mention it here.
L99. How were the crabs captured?
L102. What is the water source, salinity?
L105. What was the light source and lighting schedule?
L106. Provide the units for carapace width that you used in the models.
L106-110. You only need to define the three classes that occurred in this study. Define clearly the leg lack variable that you used in the models.
L112. Are melamine sponges standard in terms of density, color etc. or should you identify the brand and color?
L115. The relationship of the tank and cage is not clear. Was the cage floating in the holding tank referred to above? What was the cage made of? Would conditions inside the cage be similar to those described for the holding tank?
L116. Were the sponges negatively buoyant so that they rested on the bottom of the cage?
L116. What time of day were the sponges introduced?
L118. Are the crabs nocturnal?
L119. Be explicit that you recorded the latency to produce a cap as the number of days from the day on which the trial started to the morning on which a cap was found.
L120. The entire study is the experiment. When referring to a single example, it should be called a trial (‘we stopped the trial’).
L122. Move the sentence about drying the sponges to follow the sentence about collecting the sponges.
L122. Be explicit about what variables you measured to become the independent variables in your model, including the units.
L126. What equipment was used for the video recording? Was the recording made in the dark or light? What was the light source? What information did you extract? Did you take these data from all 10 recordings or only some?
L127. Provide aquarium dimensions in the order length, width and height and write L x W x H) after the last measurement to make it clear to readers.
L127. Were the crabs active only during the night?
L132-133. I suggest using the simpler and more descriptive terms ‘amount of sponge removed by cutting’, ‘size of cavity’ and ‘latency to produce a cap’. For the subheads, you can refer to ‘cutting’, ‘cavity size’, and ‘latency’.
L133. I think you need to identify what kind of models you used.
L133. You have identified the dependent variables but not the independent variables in your models. To make the statistical overview complete, briefly indicate the predictors even though your provide more details in the following subsections.
L161. Do you mean that you were unable to collect repeated data from some of the 8 animals selected for this treatment or are you reminding the reader that for 30 animals you recorded only a single choice?
L161. Should parameters be plural? It appears to be a single parameter, but I may not fully understand. This is also an issue in subsequent sub-sections.
L163. Check this term. I am not clear what you mean by ‘subjected to’. This is also an issue in subsequent sub-sections.
L196ff. Please be sure that your terms for behavior here are consistent with your use elsewhere in the manuscript.
L200. Because you make a point in the Discussion about the speed of cutting, it might be good to give an idea of the range as well as the mean and whether there was any trend for the two crabs to differ.
L201. Introducing the word ‘removing’ without a noun is confusing. Do you mean they finished cutting?
L201. Complete information. I think you mean ‘in nine of ten trials’.
L206. Explain more fully how they used their body to dig. Is the front edge of the carapace sharp-edged? Was the body moved laterally or dorso-ventrally?
L207. You refer to the night but have not told the reader whether the crabs are nocturnal and whether you recorded at night or in the day or both.
L207. Do you mean only that the behavior of rotating the body to dig with the carapace was repeated multiple times or that the behavior of removing pieces with the chelae was repeated too?
L255-265. You have not studied adaptive significance. This paragraph is not related to your manuscript. You should start with the description of the behavior (as in the second paragraph), and the heading should be something like ‘Behavioral observations’.
L266-269. Be specific about what is similar. What is the evidence that they process natural and artificial sponges similarly? You used only artificial ones.
L270. Do you mean to cut the sponge or to make it? You stated above that the average time for cutting was 50 min but that the latency to complete the cap was longer. If the average is 50 min, I would expect that the previous observation of a range of 30 – 45 min would overlap and also that this might be based on a generalization from the cited author rather than actual measurements. Can you really make the case that there is a species difference? If there is a difference between studies, could it be related to the thickness and material of the sponge rather than a species difference in speed?
L271. What was the pattern for C. hilgendorfi?
L272-278. I think that this information came from your broader study and not from the film. You should discuss latency when you discuss the modeling. You should keep the Discussion in the same order as the Methods and Results.
L277. It is not clear what you mean by time and risk sensitivity and how it relates to your findings.
L279. This is an unclear sub-heading that does not focus directly on the questions your study can address. Consider something like ‘Effects of size on sponge selection’.
L280-294. This section needs more information and better organization. See comments above.
L296-318. See my comments above.
L320. In the Fig. 1 caption, specify which parts of the figure provide dorsal and ventral views.
L332. Abandon implies that a choice was made and subsequently the chosen sponge was not used (abandoned). In this case, you are referring to choice of no sponge.
L336. Why 10 samples? Not clear what you mean. Should you have explained this in the Methods?
L344. It is not clear what you mean by ‘remarkably’. Is your intention to imply that the difference is very strong or unexpected?
L355. Your meaning with regard to mismatches is not clear.
L362. Incorrect panel description. Fig. 6A is a diagram, not the hierarchical model.
L365-366. It seems to me that assumptions of the model should be in Methods or Discussion, not a figure caption.
The References are generally well done, but books and book excerpts do not follow the PeerJ examples because they omitted the city of publication. Journal articles do not require the number in addition to the volume.
The supplementary videos are very helpful, but they need a complete description: when and where they were filmed (date, day or night), perhaps equipment and lighting, species and size of crab, speed of presentation. It appears that some of the film is in time lapse and others may be at or close to natural speed; please explain this in the captions.
[# PeerJ Staff Note: The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title) #]
Cap making by sponge crab: general and specific comments
Needs some English corrections. FAR TOO LONG! Figure captions need to be shorter.
The actual modeling of the paper is a novel approach and worthy of publication. But the paper is far too long and contains many references and comparisons that are not pertinent to the study. In the Introduction, I would delete all the references to decoration in families other than the Dromiidae. You could say that the means of decorating in the Majoidea (now split into several families) differs because they attach materials to setae instead of using the hind legs to carry their covering. The reader is lost in all of the references.
Line 71: research spelling; line 72: in field research (research is both singular and plural)
Line 83: we investigated the sponge crab Lauridromia dehaani
The last paragraph on page 2 should be incorporated into the methods. Line 85: We modeled four variates
Line 112: What is melamine and why did you use sponges of this material? ( Or is this the common ingredient of artificial household sponges?)
Line 132 and the rest of the manuscript: the depression in the sponge cap should be called a "cavity", not a "hole". A hole is considered to be deeper than a cavity.
Line 200: a groove
Line 206: I would say "excavating" instead of digging; to make the cavity larger...
Line 210: I think you should omit the line starting at "We will describe" because you say this later.
Line 240: Cap cavity
Line 257: I would confine the discussion to dromiid crabs. Homolids belong to a different family. The sponge crabs that carry shells as mentioned in line 260 do not belong to Lauridomia or related genera.
Line 261: Bedini et al. (2003) expected (do not repeat the reference)
Line 280: Few marine animals show a camouflaging behavior (decorating is used in reference to the Majoidea, not other crabs). I would delete the lines referring to Oregonia gracilis: there are alternate explanations for why big male crabs don't decorate.
Line 284: crabs were larger
I would delete the first paragraph under "Integrated extended body". Most of this material is not relevant.
Line 309: I would delete the first line. Guinot and Wicksten did not consider camouflaging behavior to be tool use. Indeed, much of the camouflaging behavior of majoid crabs consists of an innate behavioral chain with little evidence of any modification.
The figure captions are FAR TOO LONG! For example: Figure 2: Cap making behavior showing cutting, digging and carrying. A: crab grasps side of sponge, tears pieces to make groove, rotates. B-C: Crab carries artificial sponge cap; B frontal view, C lateral view. (Or something like this).
Line 451 on: be consistent in use of capitals: MUSORSTOM is the abbreviation of an expedition series, but the journal titles should not be in all capitals.
Experimental design is OK as is.
Interesting use of a statistical model.
Try to confine your writing directly to your work on sponge crabs and point out the pioneering use of your modeling technique. Your writing should not repeat itself and not contain references that are tangential to your study.
See General Comments
No comment
No comment
General Comments:
=================
NOTE: My background is in Statistics rather than behavioral studies, and the comments below are limited to my area of experize.
This will be a very useful paper in this particular area because it highlights the need and provides methods to handle grouped data. Moreover it demonstrates the perils of failing to use marginalized likelihood for model comparison. However, I feel that the coverage of this lacks a clear example (see point 3 below), and so will not be meaningful to readers who are not already familiar with the concepts of conditional vs marginal likelihood.
The English grammar is generally good for non-native English speakers, though there were several areas where it caused confusion - these have been listed below.
Specific Comments:
==================
1) There are numerous places where grammar and exposition need improving. A few examples include:
- line 16: From the wording "...including repeated samples, from the same individual crab." I got the initial impression that this study only looked at a single crab. Change to something like "...including repeated samples, from each crab."
- line 19. Change "...with the clustered data." to "...with clustered data."
- line 18. Change "...in statistical modeling. This is because..." to "...in statistical modeling because...".
- line 32. Change "...and studied the individuality in..." to "and studied individuality in..."
- line 65. Add "do", i.e., "...do we need..."
- line 71. Remove the first "the".
- line 84. Replace "one" by "each" or "an".
- line 137. Should "The..." be "We..."?
- lines 150-151. Please reword the sentence beginning "We did in the same way...".
- line 158. The phrase "...and unexpectedly abandoned the choice itself" should be reworded (to say that the crab made no choice).
2) The paragraph spanning lines 50-70 would benefit from a re-write:
- line 52: It would be useful to make it clear that the clustered data is refering to the repeating measurements on an individual.
- lines 61-63. This sentence talks about not being able to appoximate the posterior, and so needing a Bayesian approach. This makes no sense, since if one is attempting to appoximate the posterior then one is already using a Bayesian approach.
3) The second paragraph of the Statistical modeling section does a good job of talking about marginal-level WAIC . However it is not going to be intelligible to many readers without a concrete example. I would suggest that the distinction between conditional and marginal likelihood be made explicit in one of the ensuing models on pages 4 and 5. That is, give the form of (or equation of) the conditional likelihood, and of the marginal likelihood. In fact, the explanation in the above-mentioned paragraph might be better placed by moving it to a position after these explicit examples.
4) There are no citations for the methodology presented on pages 4 and 5, so it is not clear whether these types of model are entirely novel to behavioral studies (which I suspect not, since this is not the focus of the paper), or whether it is the formulation and methods to include repeated measures that is novel.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.