免費資源 · Open resources

常見問題：從研究到發表

不需要登入、不需要報名。這些是新思惟公開問答頁裡最常被參考的問題，英文版由中文原文翻譯。

31 articles

After you submitThree months with no news from the journal: should I write to the editor?Usually no. If you must, wait 1.5 times the journal's typical turnaround, and write to offer help, not to complain.
Replying to reviewersA reviewer asks for sample size and power in my retrospective study. Can I ignore it?No. Do the calculation with an online tool, put it in a table in your reply letter, and keep the tone polite.
After you submitIf a journal rejects my paper after review and offers to transfer it, is portable peer review a good deal?Weigh two things: whether the receiving journal has the impact factor you need, and whether it charges an article processing charge. You can also do your own "portable review" inside the same publisher family by disclosing the earlier rejection and answering it.
Data and statisticsWhen can I use a median (IQR) in place of a mean ± SD? Or is it always acceptable?Use the median as the mean, divide the IQR by 1.35 to estimate the SD when the data are roughly normal, or let an online calculator do it for you.
Study designHow do I write an original article from scratch?The two hard steps are making the study feasible and getting it accepted. Search the literature thoroughly, plan around the resources you actually have, and run the numbers and figures before you write the text.
Study designMy mentor has a new treatment, but the series is single-arm and small. How can I write it up?If the treatment is new and the field is interested, frame it as a feasibility study or early experience: stress feasibility and safety, and compare outcomes with the literature in the Discussion.
Study designWill journals accept a meta-analysis of case series or observational studies? What if every study is single-arm?Yes. Pooling controlled studies is valued more, but when single-arm studies are all your field has, pooling them is meaningful and publishable, especially if no one has done it yet.
Data and statisticsData collection is done. Which MedCalc functions should I use to build Table 1?Decide the story first; Table 1 is built by hand to serve it. Report descriptive statistics for each group side by side, and leave out significance tests of baseline differences.
Data and statisticsOne trial has three treatment arms and one control. How do I extract it for my meta-analysis in CMA?Either drop the arms that are not relevant to your comparison, or combine the eligible arms into a single group, as the Cochrane Handbook recommends. Combine SDs with the Handbook's formula, which uses N − 1.
Replying to reviewersA reviewer wants 95% CIs. Should I report mean, SD and 95% CI, or drop the SD?Either is fine. I report all three: it satisfies the reviewer, and future meta-analysts will not need to convert back to the SD. For small groups, the CI should use the t distribution.
Data and statisticsTwo observers measured the same thing. Should I compare them with a paired t test?This is inter-observer variability, and the method is a Bland-Altman plot. Repeated measurements by the same person are intra-observer variability.
Data and statisticsMy meta-analysis was rejected with confusing comments about SMD and Jadad scoring. And how do I switch to SMD or Hedges' g in CMA?A difference in means works when all studies share a scale; otherwise use the SMD or Hedges' g. For randomized trials, reviewers increasingly expect the Cochrane RoB 2 tool rather than Jadad scoring.
Data and statisticsHow many studies does a network meta-analysis need, and how many per treatment?For a first network meta-analysis, about ten trials is a good balance, with five or six as a minimum. A treatment can enter the network with a single trial.
Data and statisticsBoth groups received identical-looking pills. Is that allocation concealment?No. Identical pills are about blinding. Allocation concealment means nobody enrolling patients can know the next assignment, for example with opaque sealed envelopes or central randomization. Blocks filled in order of enrollment are still random if the order within each block was random.
Data and statisticsOne of my subgroups has a Hedges' g of 1.1. Is that too high?It is large but not abnormal. Conversion errors usually produce values of 3 or 4. If you have checked your extraction, you have done your part.
Replying to reviewersA reviewer wants me to drop one high-risk study from my six-study meta-analysis. Is that reasonable?Don't remove a study after the fact without a rule. Keep the original analysis, and show the reviewer a sensitivity analysis without that study or an analysis by risk of bias. If you do change the analysis, report it as a change from your protocol.
Data and statisticsSome studies used a paired t test and others an unpaired t test. Can I pool them?A before-and-after comparison in the same patients and a comparison between two groups are different quantities. Crossover trials can be pooled with parallel-group trials using the paired effect estimate, ideally as a separate subgroup.
Data and statisticsNot every study reports an SD. Can I use studies that give a 95% or 90% CI instead?Yes. Convert the CI to an SD by hand (the width times √n, divided by 3.92 for a 95% CI or 3.29 for a 90% CI; use t values for small groups), then enter it into CMA. The Cochrane Handbook gives the formulas and is easy to cite.
Data and statisticsPooling studies that measure one outcome on different scales: Cohen's d or Hedges' g?Hedges' g, especially with small samples.
Writing the paperHow do I know whether my work suits a review article? Do I need my own original articles on the topic?Unsolicited reviews are accepted more readily when you have published in the field, when your clinical work and images are outstanding, or when you see a pattern others missed and quantify it with a systematic review and meta-analysis.
Writing the paperHow do I choose a topic worth writing a review article on?Choose a field where you already have a track record, a topic residents would want to download on sight, or one where you have a distinctive view of your own.
Writing the paperMy target journal asks for "How this fits in" and "Implications". What is the difference?Read recent issues to see how other authors wrote these sections, and follow their angle. Each section of a paper has its own purpose, even when two sections look alike.
Writing the paperI found an error in an earlier study. Should I point it out?My advice to early-career researchers: be gentle. Prefer a courteous query, don't disparage earlier authors in your Discussion, and in an updated meta-analysis simply extract the correct numbers.
After you submitAfter publication, a letter to the editor says related studies already exist. How should I reply?Thank the writers, explain how the studies they cite differ, give the dates of those published after your submission, and add any further studies you found yourself.
After you submitThe journal rejected my paper but invited me to resubmit after revision. Why not just give me a major revision?Reject-and-resubmit has become common. Revise, resubmit quickly, and say in the cover letter that the editor invited it and what you changed.
After you submitMy paper was accepted. What should I watch for when checking the proofs?Check every word. Typesetting errors in affiliations, author names, text and references are common.
After you submitHow does desk review work? When does a paper go to peer review, and what does an "undecided" status mean?From my experience reviewing for traditional journals: some papers go straight to reviewers, while others wait in an unclear pool. An "undecided" status is still much better than a rejection.
After you submitMy open access paper was accepted, and I can choose CC BY or CC BY-NC. Which should I pick?If the price is the same, it depends on how you feel about commercial reuse. I lean toward CC BY.
Replying to reviewersDoes a major revision mean my paper is likely to be accepted? How do I answer comments that would change the paper a lot or conflict with my design?In my experience, most major revisions end in acceptance when authors identify the reviewer's two or three core concerns and answer them with respect and evidence.
Research ethicsIf I split the data from one project into two papers, does the second paper need to mention the first?Yes: cite it, and say the data come from a larger project. Don't copy text between the two, and make sure each paper asks a distinct question, or you risk being criticized for salami slicing.
After you submitHow do I cope with anxiety while my manuscript is under review?Start the next paper. Writing the next one, perhaps of a different type, turns your attention to productive work, and news on the first one tends to arrive while you are busy.