This blog post is the third part of the statistics theme. It’s easier if you read the previous posts first.
Sometimes, the statistical analyses are like a separate part of the study. This may be the case when someone else than the main author has done and written the statistical analysis. Usually, the Results section is better without a subtitle ‘Statistical analysis’.
It would be best to present both i) the actual values observed (measured) and ii) the statistical significance. In general, it is better to mix these two than to present them separately.
There are roughly two ways to write the results: 1) the focus is in the values observed, 2) the focus is in the significance and statistical analysis values. In other words: 1) starting from the values observed, 2) starting from the statistical analysis.
The difference is not great if both the actual values observed and statistical values, such as p values are presented. The problem will occur when either of them is not presented of when they are presented in separation from each other.
When the results have been written from purely a statistical analysis point of view, a reader has difficulties in assessing the actual results. Readers should always be able to see also the actual values observed, not only the statistical values. Readers should also see the variation around the means, so give the treatment means, SDs or SEs, and n.
How to mix the statistics and the values observed in practice? Try to start writing the values observed and then add between the statistics to show whether the results were significant or not. In other words, write first about the treatments and add the significance. Which of the treatments had the highest value and was it significantly different from the other treatments?
Think carefully what p values you should present. If you want to present many statistical analyses details, like F or MS values in ANOVA, make a table of them. In most cases, p values are enough.
However, the journal instructions about presenting the statistics vary. Some journals want details more, some less. Always read the instructions to authors and follow them. For any journal, however, ‘the more p and F values the better’ does not hold true – at least I hope so.
This blog post is the second part of the statistics theme. It’s easier if you read the previous post first.
It is not so simple that any statistics would always help in the acceptance. The statistics may be too simple to deserve much space. In a scientific article, the statistical analyses should not be presented in such a detailed way like in a MSc thesis. All general method explanations, for instance, about ANOVA, are textbook knowledge, and not wanted in a scientific article. If the analysis is only little used, and you think that your readers do not know it, then the idea of the analysis is good to explain in the methods.
ANOVA and post-hoc tests do not belong to the analysis needing any explanation, not in any case. You should explain the design, and the analysis should fit into the design. In a simple case, it is enough to write that you had a completely randomized design, and you performed x-way ANOVA followed by Tukey’s test. That is all you need in the Methods text.
Some journals want the exact p values rather than one limit, such as p < 0.05. Some journals want also the ANOVA table. If that is written in the instructions to authors, it is best to follow this instruction. In many cases, however, you can well use the p value 0.05 as the limit for statistical significance.
The importance of the actual p value is not so self-evident in ecological research where the understanding of the processes and mechanisms are in focus. My opinion is that in the ecological research, which I was doing, the actual p values with three of more digits do not give any additional information to the research presentation; most importantly, they do not bring anything to the assessment of the ecological importance, ecological meaning of the results. The worst case is that you are presenting the p and F values with five or more digits, and do not say anything about the meaning of the results in practice.
From one article draft reporting a relatively simple experiment, I counted 16 p values varying from p = 0.00000000149 to p = 0.02692. Unfortunately, also values, such as p > 0.25464 were suggested to be published. Do you understand what the problem with values is? If not, please ask me.
If you do not have any statistical analysis in the article, it is an easy task for a reviewer to reject the article. You should have thought of the analysis before you started to do the research, but in case you did not, now it is the time to think.
The basis of the scientific research is that you should be able to say how probable it is that your results are true also elsewhere or maybe in the next year. This means that the result is not random, but it is real and reliable.
Although statistical analyses may not be so extremely important in the development of ecological sciences, you still, in practice, must do some analyses. Although you think that the results can be seen without any analysis, in practice, you must do at least one analysis.
Because the peer-review system is many times arbitrary, you cannot know in forehand what the reaction of the reviewer will be. However, one thing is almost sure. The reviewers want to see at least one statistical analysis. It does not need to be any complicated analysis, most often reviewers are satisfied with a simple analysis of variance, correlation analysis or principal component analysis.
Researchers in ecology are often not extremely interested in statistics, and they accept any statistics they see. However, they know that the scientific style is that some analysis should be done. That is why you better have at least one statistical analysis or some kind of a data-analysis in the article. Presenting the variation around the treatment means with SD or SE is a must, but it is not enough.
You really can improve the possibilities to get the article accepted with a statistical analysis – even if the statistics really do not improve the research nor its presentation. I know many examples, where the value of an article has been managed to be increased with complicatedly looking statistics. It certainly is possible to get a poor data accepted with complicated statistics. However, I am not recommending it here. I recommend that you plan and perform experiments with care and get good data and finally, do simple and common analyses that most people can understand.
In general, the statistical analyses are performed to show if the values observed are statistically significant. In many cases, in practice, the significance test helps readers to see the result easily without reading and comparing all values so carefully. The statistical significance tests help to write the results, in practice.
The tests also help authors to assess if the values or differences matter. However, the significance test itself does not tell about the ecological importance. The ecological meaning and relevance should be done separately from the statistical testing, but as a complex issue, it is a theme of another post.
There are many essay writing services available. First of all, I want to say that we in EcoSciEdit and essay writing services are not competitors. We have different tasks and aims. Essay writing services promise to write the whole stories, from research articles to the thesis, by themselves. Using this kind of a service may fit to students or other persons not aiming to be researchers themselves. However, it is not any good system for students or anyone aiming to achieve a researcher career, in my opinion.
This is because by writing the articles yourself, you can develop yourself to be a better researcher. By reading what the other researchers have been doing, you will learn to do better research yourself. Only by knowing what is going on in the science of your research area, you can find out the research questions that would be interesting to the international audience.
When you build your next study around this novel question, you will get it easier accepted by higher-quality journals. Although good writing helps in publishing, doing science is not only writing. Science is, of course, having relevant research questions and answers to them. It is really several years of hard work to be on that high level in science. I believe that almost no one is at that level after finishing a PhD.
So, our EcoSciEdit editing and writing service is not competing with essay writing services. We aim to teach you writing. Our aim is that you do not need our service for ever. By using an essay writing service, you will need this service for ever. And most importantly, without reading a lot of articles already published, you will never find out the correct research questions that the international scientific journals are eager to publish.
In fact, I do not understand how someone not familiar to the research area in question is able to write a Discussion. An editor can write something, of course, but I do not believe that it is of reasonable scientific quality. Or it can be of reasonable quality, but in that situation, it needs much work. Writing even a short discussion needs many days of work, because you must search for articles and read many of them and compare your results to previously published. Therefore, I do not understand, how a Discussion is possible to be written by someone outsider with a cheap price. Anyone knowing that, please tell me.
I think that the system where someone else than the researchers themselves write the articles does not fit to the science, and that is why I do not act as a 'ghost writer'. I want to develop the manuscripts together with the authors. I believe that the authors can learn to write themselves in the long run and start to write better articles in the future.
A PlosOne review process of one article is worth of sharing. I do not remember the times taken correctly, but all this took very long, several months regardless of continuous asking about the process.
The editor told to each inquiry that he had not found any reviewers. It is good to mention that the manuscript dealt with a rare subject. Even so rare that during the writing, it had been difficult to find ‘enough’ relevant articles to be referred. However, on the other hand, the subject was pretty general and the methods were relatively simple microbiology and, in my understanding, it should not have been so extremely difficult to find any reviewers.
Finally, the reviews came. There were three of those, and none of them suggested rejection. Each of them suggested revision, one minor, and the two other reviewers wrote that the topic is important and emerging. Two latter reviews were of high quality and the reviewers seemed to know about the subject. However, the editor rejected the manuscript. The editor required additional experiments, although he should have understood that this was a field study and not possible to make any new experiments. None of the reviewers suggested more experiments, they suggested minor-moderate changes in writing.
The oddest thing to happen was still to come. It appeared later, that there had been one more review that had been very positive. We came to know about this review, because the reviewer in question contacted the author and told about it. This reviewer had been studying the same research subject. We had no reason to believe that this positive review would not have been on the editor’s table.
The only explanation I can imagine for this is that because the subject was really rare, it could have been presumed that it will gain only few citations in the near future. PlosOne is a journal wanting to increase its impact factor (although all journals want, of course) and therefore, it may keep the possibility to gain citations as one important principle to assess the manuscripts.
Finally, the manuscript was pretty fast accepted elsewhere.
If you want to share your review experiences, go to SciRev sites and write them there. You can also compare different journals on the sites and read experiences.