BMC科研

1天前

原始研究成果的交流，要求既能有效地表述研究人员希望传达的事实，同时也能表达作者想要提出的设想。第一作者需要将所有其他作者的想法组合起来，形成一篇连贯的、有价值的和高效的论文。

您首先需要慎重考虑并确定自己希望表达的信息。我们建议您在确认自己已完整地、有逻辑并有效地表述了这些信息，并且删除了任何对传递这些最重要的信息无关的句子和段落之后，再提交您的论文进行润色。

在本篇，我会提供18个想法和建议。综合起来，这些建议应该能提高您在论文中表述原始研究所得重要成果的能力。现在，在这第一篇博文里，我将指出一般注意事项，以及关于论文的方法和结果部分的一些详细说明，希望您能阅读这两部分并从中受益。

1.论文的目的不应该是发表点东西来填充简历，以增加在学术界被提拔的机会。而是为了传达一些重要的科学发现，而在此基础上，其他人可以使用它们来改善人类的生活条件。一篇创作良好的论文不仅可以让激动人心的新颖发现获得关注，也可以避免不必要的重复研究，还可以告知别人应该避免的错误。只有在值得写并可以帮助别人推动科学发展时，才应该开始创作论文。

2.浓缩的总是精华。写一篇凝聚了你想要表达的信息的好短文，比起写一篇充满废话的长篇大论相比，更费时。但是花些时间来去除废话保留精华还是值得的。

3.如果一项研究是假说驱动的，那么就请在引言部分提出假设，可能的话也写在摘要里面。然后确保论文的每一部分在某种程度上都与假说的检验过程或讨论相关。如果哪一部分与假说无关，那么就删掉它。如果它与假说有关系，则要确保能明确告知读者它们是如何相关的。

4.如果研究发现了假设之外的意想不到的结果，那么在引言部分先承认这一点，读者就会知道你所展示的是一个偶然的发现。如果最初的实验设计并非是用来观察这些意料之外的结果的,读者和审稿人则不会介意研究欠于完美之处。

5.在任何研究开始之前,确保方案设计可以很自信地回答所假设的问题。如果研究设计有缺陷，无论写作和编辑如何完美，仍然会是一篇不佳的论文。

6.如果一项研究不是假说驱动的，一定要记住这一点。然后在论文创作时传达出这一信息。

7.在你开始论文(假设是一个假说驱动的项目)写作时，写下你的假设。假设是锚，将阻止你的思想偏离主题。接下来，列出你想向读者传达的信息。所有信息都应该与假设相关，尽管他们并不需要都支持它。按重要性或其他逻辑顺序将这些信息排列好。你想要读者最清晰地记得你做的所有工作中的哪几项？这些就是你的重要信息。在你的列表中的每项重要信息，都将成为稍后讨论部分中每个段落的首句，而且段落的顺序就按你列表中的顺序即可（这将在接下来的博文中谈论到)。通常一篇论文会有2至5个关键信息点，当然也可多可少。

8.我不建议按照论文各部分最终的顺序撰写论文。首先应该编写材料与方法部分，提供所有必要的细节以便他人能重复此研究。写作方法部分的每一段落时，重读你的假设并回忆一下你为什么要做这些实验。如果你都不清楚这些实验与你的假设之间的联系，那么读者也很难看出来。稍后可以对方法部分进行再修饰。

9.通常可以在描述每个方法的开头，增加一些几个单词组成的短语，用来解释为什么采用这个方法，以及它如何与项目中的其他实验相关联。例如，“为了确证mRNA方面的结果，我们使用……分析了生成的蛋白质”或“为了进一步评估心理测试在ADDH诊断中的有效性，我们通过采用……进行了长期效果评估”。

10.统计方法。许多读者会直接相信作者能确保其研究使用了适当的统计检验进行分析。因此，对于作者来说，选择正确的统计分析方法并理解它们，是很重要的。在必要时可以寻求统计学家的帮助。

11.接下来撰写结果部分，段落之间的顺序应合乎逻辑。重读你的假设。将每一个你要报道的发现与要检验的假设联系起来。

12.在结果部分，有时会对实验结果进行解释，但更常见的是将这些解释留到讨论部分。如果你想在结果部分解释你的发现，那么你必须只表述事实，并且在讨论部分不再重复。应该避免在结果部分进行猜测。猜测（使用单词如“也许”、“或许”、“大概”、“可能”、“有可能”）要留待讨论部分进行。

13.在结果部分，尽可能以高效的方式呈现事实。如果一个表或图能最简单地表述结果，那就采用表和图。完全可以这样表述:“血清实验室化验的结果如表1所示”。也可以添加这样的句子:“请注意在12个化验分析物中有5个显著升高。”对读者来说，比起试图破译一段冗长的包含数据的文字来，表或图更容易看懂。如果有出色的表和图，结果部分可能也就几句话。

14.数据，不管是在文本还是表中，都不应该包含小数点后任何毫无意义的数字。从本质上讲，如果在现实世界中并不是真正重要的，那就不要提供小数点后的数字。比较456.237 +/- 64.243和234.549 +/- 76.345，是很少有意义的！使用456 vs 234就好，不管检测的是什么变量。实际上，使用过度和毫无意义的数字，会使审稿人认为作者并未考虑数据的实际意义，而只关注数学和统计。

15.避免屈服于p值。我将用一个极端的例子进行说明。如果p值是0.049，你会认为治疗是成功的，而如果p值0.051则认为治疗完全失败吗？你不应该这么认为。在现实中（纯粹的统计意义之外），对于p值为0.049和p值为0.051，本质上是没有真正的区别的。我们不应以p值落在随机值0.05的哪一侧而得出质变性的结论。0.05不是一个神奇的数字。这就是为什么p值应该被写成p = ###，而不是p<0.05或p>0.05。确切的数字，可以让读者看到均值相互之间到底有何差异。同时，使用大量的零并非经常管用。使用p<0.0000001，在大多数情况下都看起来有点傻，除非是在分析成千上万个变量时。几乎在所有的论文中，使用p<0.001就已经足够表明数据之间的差异可能代表真实世界的差异。

16.图和表。想想看读者想要看到什么。将标题和轴标签制作清楚。因为图表可能会压缩，因此尽量使用大的字体。点图往往比柱状图更加信息化，尤其对于医学文章而言。点图能展示个体数据，而患者都是个体。基于研究中的均值数据柱状图对患者进行临床管理的例子是非常罕见的。均值不能给出全部真相，因为个体很少等同于均值。

17.图例应该传达出结果及其说明，读者无需阅读论文即可理解。一个图的图例的第一部分应该表述你想通过图传达的主要信息。一些期刊不希望图例中包含解释，但大多数不会介意。

18.现在回到方法部分，确保结果部分报道的任何一点都有相应的方法。加入之前忘记的内容到方法部分，同时删除你发现事实上并非包含在这组实验里的那些方法。后面这种情况通常并不多见，但我们偶尔会发现方法部分的某个段落所述方法，实际上在整个研究中并未使用过。

在这个两部分博客系列的第2部分中，我们将对如何撰写非常重要的引言、讨论、和摘要，提出想法和建议。谢谢！

英文原文如下：

Manuscript creation for English language journals. Part 1 of 2.

Communicating original research findings well requires efficiently telling the factual story that the researchers wish to convey, as well as the speculations that the authors wish to propose. It is the first author who needs to organize the thinking of all the other authors into a manuscript that is coherent, worthwhile, and efficient.

We recommend submitting for English editing after you are comfortable that you have fully, logically, and efficiently presented the messages that you have carefully decided that you wish to present, and after deleting any sentences and paragraphs that are not essential to conveying the most important messages.

In this blog post and the one that follows, I offer 30 thoughts and suggestions that, taken together, should improve the ability of your manuscript to communicate the important results of your original research efforts. The second blog postwill focus on the Introduction, Discussion, and Abstract. For now, this first blog post presents general considerations and specific comments regarding the Methods and theResults section of a manuscript. I hope you will read them both and that you and your manuscripts benefit from them.

1. The purpose of a manuscript should not to be to get something published in order to fill ones curriculum vitae and improve chances of being promoted within academia. Instead, a manuscript’s purpose is to convey important scientific findings in a manner that others can use to improve the human condition. A well-written manuscript can not only gain attention for exciting novel findings, but it can also prevent wasteful repetition of research efforts, as well as inform others of mistakes to avoid. One should only write manuscripts that are worth writing and that will help others move science forward.

2. Shorter is always better. It takes more time to write a good short article focused on your message than it does to write a long article filled with noise. But it is worth the time to remove the noise and leave the message concisely stated.

3. If the study was hypothesis driven, state the hypothesis in the introduction, and possibly in the abstract. Then make sure that everything you put in the manuscript is in some way contributing to the process of testing of discussing that hypothesis. If it doesn’t relate to the hypothesis, remove it. If it does relate to the hypothesis, make sure the reader is told exactly how it relates.

4. If the study identified unexpected findings outside the hypothesis, then admit this in the introduction so that the readers will know it is a serendipitous finding that you are presenting. Readers and reviewers will forgive a less than perfect study when it was not originally designed to examine the unexpected observation.

5. Before embarking on any research, make sure the study design can confidently answer the questions of the hypothesis. If the study design is flawed, the manuscript will be poor no matter how well it is written and edited.

6. If the study is not hypothesis driven, make that clear in your mind.And then also convey that message in the manuscript.

7. Before you start writing the manuscript (assuming a hypothesis-driven project), write down your hypothesis. The hypothesis is the anchor that will keep your mind from drifting off topic. Next, make a list of the messages you wish to convey to the reader. All the messages should relate to the hypothesis, although they don’t all have to support it. Organize the list in order of importance or in some other logical order. What do you want the reader to most confidently remember from all the work you have done? Those are your important messages. Each important message in your list will then serve as the first sentence of sequential paragraphs in the Discussion that you will write later (and that will be discussed in the next blog post). Often 2-5 key messages are identified, but there can be more or less.

8. I do not recommend writing the manuscript in the order it will be read. Instead, write the Methods section first, with all the detail that is necessary for others to replicate the study. For each paragraph of the Methods section, re-read your hypothesis and remember why you did the experiment. If you are not thinking about the connections to the hypothesis, it will be difficult for the reader to do so. You can fine tune your Methods section later.

9. It is often helpful to add a phrase to each Method description consisting of a few words explaining why this method was performed and how it links to other experiments in the project. For example, “To confirm our mRNA evidence, we analyzed protein products using…..” Or, “To further evaluate the utility of psychometric testing in ADDH diagnostics, we performed long-term outcomes assessment by means of…”.

10. Statistical methods. Many readers will simply trust the authors to make sure that the appropriate statistical tests are used for analysis. As such it is important that the authors choose the correct statistical analytic methods and understand them. Getting help from a statistician is wise.

11. Next, prepare to write the Results section in a sequence of paragraphs that are entirely logical. Reread your hypothesis. Connect in writing each reported finding to the hypothesis that you are testing.

12. Interpretation of the results of the experiments within the Results section is sometimes done, but more often is reserved for the Discussion.If you do explain your findings in the results section, you have to stick to FACTS only, and don’t then repeat yourself in the discussion. Refrain from speculating in the Results section. Speculation (using words such as “maybe”, “perhaps”, “possibly”, “might”, “potentially”) is reserved for the Discussion.

13. In the Results section, present the facts in as efficient a manner as possible. If a table or figure is the easiest way to state the results, rely on the table and figure. It is totally acceptable to state, “The results of our serum lab assays are shown in Table 1.” It is also okay to add to this, “Note the marked elevations in 5 out of the 12 analytes assayed.” It is much easier for a reader to examine the table or figure than to try to decipher a kilometer long paragraph full of data in the form of prose. Results sections may have very few words if there are excellent tables and figures.

14. Data—both in the text and in tables—should be presented without any insignificant digits. Essentially, don’t provide numbers after the decimal if those numbers are not truly important in the real world. There is very little point to comparing 456.237 +/- 64.243 with 234.549 +/- 76.345! 456 vs 234 is just fine, no matter what variable is being examined. Indeed, using excessive and insignificant digits is an indicator to reviewers that the authors are not thinking about the real meaning of their data, but rather are stuck focused on mathematics and statistics.

15. Avoid being tyrannized by p values. I will use an extreme example.Would you say that an intervention was successful if the p value is 0.049, but have considered it a complete failure if the p value had been 0.051? You shouldn’t. In reality (outside of arbitrary statistics) there is essentially no true difference between a p value of 0.049 and a p value of 0.051. And we should make no leaping conclusions based on which side of an arbitrary value of 0.05 that the p value falls. 0.05 is not a magic number. This is one reason why p values should always be reported as p=###, instead of just p<0.05 or p>0.05. The actual numbers provide the reader with an opportunity to see how likely the means are actually different from each other. Also, using a lot of zeros is not often helpful. p<0.0000001 looks silly in most cases, unless there are thousands of variables being considered in one analysis. p<0.001 is sufficient in almost all typical manuscripts to convey the message that differences in the data set are likely to result from actual reality.

16. Figures and Tables. Think about what a reader would want to see.Craft titles and axis labels clearly. Make font sizes as large as possible because the graphs will be shrunk. Recognize that dot plots—particularly for medical articles—are often much more informative than bar graphs. Dot plots show individual data, and patients are individuals. It is the very rare case indeed in which a patient should be managed clinically based on a bar graph of mean data in a research study. Mean data don’t tell the whole truth, because individual people are rarely average.

17. Figure legends should convey results and interpretations, without the reader having to read the manuscript. The first part of a figure legend should state the main message you wish to convey with the figure. Some journals will not want interpretation included in figure legends, but most won’t care.

18. Now go back to the Methods section and make sure that anything you reported in the Results section has appropriate methods sections supporting it. Add to the Methods whatever you have forgotten, and delete from the Methods anything that you realize wasn’t actually part of this set of experiments. Usually this is not an issue, but we do occasionally find whole paragraphs of methods that were actually never used in the study.

In part 2 of this 2 part blog series, we will present thoughts and suggestions on writing the very important Introduction, Discussion, and Abstract. Thank you for your time so far!