Review History
A survey of researchers’ methods sharing practices and priorities

Thank you for your thoughtful revisions and responses our helpful reviewers' comments as well as my own. I am delighted to accept your very well-written manuscript and look forward to it generating discussion. Each time I re-read it, I'm left wanting to think more about the questions you raise and I hope that you will continue to work in this important area.

Thank you for your interesting and well-written manuscript. As you will see, two reviewers have looked at your work and made a number of comments for you to address. For each comment of theirs, and the additional comments I’ve added below, please provide a response explaining how you have changed your work in response to it or why you have not made changes for that particular comment.

1) I wonder if it would be helpful to use “questionnaire” when you refer to the form itself (rather than “survey instrument” on e.g. Line 75). For example, Line 81 talks about “The survey design” which could potentially refer to the questionnaire or the methods used to identify and invite potential respondents.

2) I appreciate that this point might also appear stylistic, but references to “Likert”, and particular “Likert scale” (e.g., Line 87), sometimes raise questions, including about the equal-spacing along some construct. Sometimes “ordinal item” or other options (see, for example, https://www.john-uebersax.com/stat/likert.htm for one perspective and some suggestions around naming) is useful to avoid such pedantry. I’ll leave this point entirely up to you.

3) I wonder if readers would be interested in knowing a little more about the pretesting stage (Line 95). In particular, how was this done (including who was involved) and how did this result in the questionnaire in [18] (and where did the questionnaire in [17] inform this—Lines 81–82)?

4) I was at times unclear whether you were describing the data in your sample (in which case proportions and means+SDs would be appropriate) or making estimates for the population (where CIs (or SEs) become appropriate, and p-values would be a useful addition in places). Given the survey design, I’m not presently convinced that you have a well-defined population to make inference about, but would be very happy to hear arguments to the contrary. This makes the CIs around proportions and means seem out of place to me. For the means, you could simply replace the CI half-widths with SDs; for proportions the SDs are defined in terms of the proportions. At the same time, comparing responses by groups can be enhanced by formal statistical testing, which you do at times, and I think some readers will look for non-overlapping CIs as evidence of differences (non-overlapping 95% CIs indicate statistical significance at approximately the 0.01 level and so the interpretation mentioned Reviewer #2 is more stringent than a conventional test at the 0.05 levels—note also that both Reviewers touched on the issue of multiple comparisons, which could be addressed by positioning this as an exploratory, hypothesis-generating study if that is your intention, or by more formal treatment of multiplicity, although you would need to think carefully about where you want to protect your family-wise error rate). While point-estimates, including proportions/means, can be questioned without a well-defined population and/or non-response (including non-completion here), differences are often be expected to be more robust. Two possibilities that occur to me would be to make the tables entirely descriptive (i.e., remove CIs from proportions in Tables 2–7 and replace CI for means with SDs in Table 8) or to do this and add a full set of formal statistical tests looking at comparisons of interest.

5) Related to the above, and as mentioned by both Reviewer #1 and Reviewer #2, a t-test unexpectedly appears in the Discussion (Lines 364–366) where I agree with them that the approach should have been covered in the methods and that the findings belongs in the results, but I’d like to add that actual p-values should be presented in all cases (perhaps with “p<0.001” for the smallest values). The magnitudes of these differences would seem more important than statistical significance and I think readers will appreciate seeing your interpretation about the practical significance of these results. Similar points would apply to Lines 371–376, 376–379, 385–389, 391–393, and perhaps elsewhere. Note that readers will want to know in the methods a little about the checks you used before applying these tests (do you have reasonable evidence that the populations are sufficiently normally distributed, and that variances are sufficiently equal across groups unless a Welch t-test was used? This should help to address a comment from Reviewer #1 about non-parametric alternative approaches.) Note also that a non-statistically significant finding (e.g., Line 393) warrants interpretation of the CI for the difference to see if this (in an inferential sense) rules out or fails to rule out practically important effects. This would be a reason in favour of retaining “statistically significant” and adding “of practical importance” or similar when appropriate (c.f. comment from Reviewer #1).

6) With Figures 3 and 4, I wonder if showing both public sharers and non-public sharers would help readers to appreciate the differences more than showing overall results and public sharer results. This is especially so given the discussions (e.g. Lines 362–363) comparing these groups.

**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.