swuecho

1/8/2013 - 3:24 PM

A major problem with this MS is that the aim of this study is not sufficiently clear. Is it the case that this novel statistical technique has never been applied to determining instar number in insect larvae? If yes, and your main goal is thus just to advertise this statistical approach to entomologists, say this clearly and (re)organize the whole text accordingly. However, if you would additionally like to say that 1) this method determines the number of instars correctly and/or 2) it works better than alternative approaches then I would say that, as it stands, there is no sufficient basis for either 1) or 2). To say 1), you should have used an alternative method of determining the number of instars in parallel, how could you otherwise know that your method returned the correct value? Why not to just monitor the development of a sample of larvae on individual basis and to count molts/instars? To say 2), you should have actually applied different methods, and compared the results. One way of doing so might be to progressively reduce the sample size by random omission of observations. You could then study which method ‘gives up’ earlier.

Yes, this novel statistical technique has never been applied to determining instar number in insect larvae. This method determine the number of instars correctly. This is can be verified by the publication.

Title: Der Einfluss der Nahrung auf Entwicklung, Wachstum und Prareproduktionsphase von Blaptica dubia Stal (Blaberoidea, Blaberidae). Author(s): Hintze-Podufal, C. ; Nierling, U. Source: Mitteilungen der Schweizerischen Entomologischen Gesellschaft. Volume: 59. Issue: 1-2. Pages: 177-186. Published: 1986

Actually, We thought the instar number was unknown and not one had ever studied them. This is our mistakes for not doing throughtly literature review. Thanks to your correction, we find the instar number in dubia cockroaches (*Blaptica dubia*) is the same with our analysis.

we also monitor the develepment of dubia cockroaches in our lab, it also support our analysis.

More specific comments:

- how would this method perform if there was intraspecific variability in instar number? (How) can you be sure that this was not the case in your sample?

we didn't perform analysis like that. If the insect have intraspecific variablity in instar number, the data you collect is the mix of different sets of data. As far as I understand, there is no methods available to deal with this kind of complexity.

In our case, we verify our results use the Dyar-Brooks law. Our results comply with Dyar-Brooks law. It is give us more confidence to believe there is no intraspecific varibility in dubia cockroaches. However, this is just give us more confidence. We still can not say that there is no intraspecific varibility in dubia cockroaches. We can make sure that only after we get the data from real monitor.

This may lead to another question: why can you advertise your methods if there are limitations in it? All methods have limitations. For insect instar numbers, there are so many cases, and you can not find a methods to rule them all.

For instar determination, the best way is direct observation, but many cases it is not easy, or not possible or too boring.

so, there comes brooks-dyar's rule. Dyar's rule said the growth ratio is a constant in instars. you need not observe day by day in the whole life history of an insects. you get two instars and you know the equation. if what you get is the first and the last, you know all. However, there is still problem. Many of them do not comply with dyar's rule. If you use dyar's rule, obviously you get a wrong conclusion.

In order to be more general, we use the clustering methods. the assumption is that the measurement of the insects in the same instar is more similar than the measurement between different instar. It is more general than dyar's rule. There are cases when insects molt, and they do not change in the measurement very much, so we can not distinguish . In these cases, clustering methods will fail. there are other limitation, such as intraspecific variablity and in these cases we can not get a good result. The cases dyar's rule can be used is only the subset of clustering methods can.

so the main point is as long as the measurements in a group is more similar than between groups, we can use clustering method. It is more general than Dyar's law and it is the most general methods we can find so far in instar determination, but there is still limitation (as always).

line 38 – 40 and later throughout the text. “Instar determination” not sufficiently clear, what do you mean? “Determining instar distributions” not clear either, say more clearly throughout the text.

line 70: as far as I understand, the assumption of normal distribution is nowhere in this equation; fk(x) not defined;

line 86: it appears irrelevant here how much odor do they produce; may it be understood here that the enthusiasts themselves consume the cockroaches (maybe not, my English is poor)?

The dubia cockroaches is widely used as a food source for pet. One of the reason that make them popular is that they do not produce odor as other cockroaches. When the pet lovers raise them, less odor make the cockroaches less disgusting.

- line 88: in the light of what you have just said, it is hard to believe that life history of B. dubia is unknown; actually, for example, Zoological Record knows about 24 papers on this species, including this one Title: Der Einfluss der Nahrung auf Entwicklung, Wachstum und Prareproduktionsphase von Blaptica dubia Stal (Blaberoidea, Blaberidae). Author(s): Hintze-Podufal, C. ; Nierling, U. Source: Mitteilungen der Schweizerischen Entomologischen Gesellschaft. Volume: 59. Issue: 1-2. Pages: 177-186. Published: 1986 which also contains data on instar number.

thanks to your information, we find the publication. The instar number recorded in the paper is the same with our prediction. This makes our analysis more beilievable.

the equation after line 108 seems to unnecessarily repeat the equation after line 70;

not all symbols are explained for the following equations;

Does the analysis implemented in R package “mclust” have a name? In any case, the assumptions and the purpose of this analysis should be explained more fully and more understandably; in particular (line 124) please explain the meaning of ‘hierarchical’, explain ‘the EM algorithm’. What are the strong and weak aspects of this analysis ?

mclust:

Normal Mixture Modeling for Model-Based Clustering, Classification, and Density Estimation

more specific details of statistical methods are required.

- line 128: can really BICs of particular CLUSTERS be compared? If yes, please explain what does this mean. I understand comparing BICs of different models.

I agree with you. Actually, we are comparing different sub-models. The best one is what we need.

- I do not know what are the simultaneous T-square intervals. This is my own fault, of course, but please consider helping other readers with similarly insufficient statistical background.

similar to confidence interval, but for multivariarate date.

- I sorely miss univariate distribution plots of your variables. I would like to see how clear were the seven clusters in your data: if we could easily count the 7 instars from the distributions, your work does not really prove the usefulness of your statistical method. In contrast, the bivariate plots (Fig. 1 – Fig. 3) seem to be of little use and may be omitted.

I will give you.

the bivatriabte plot give the visulation of the clusters. It is important because you can jude if your clustering is reasonable or not.

- Table 2. Regressions… what is the x-axis variable?