Impacts of Data Synthesis: A Metric for Quantifiable Data Standards and Performances
, Pekka Siirtola 1, Satu Tamminen 1, Mikael J. Knip 2,3, Riitta Veijola 4 and Juha Röning 1Round 1
Reviewer 1 Report
Congratulations on a very interesting study. I believe the analysis is relevant and your findings can help guide researchers and healthcare stakeholders in their work with identified data sets. However, the organization of the manuscript and the content in several sections could use further work. I hope my comments could help in the process.
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
1. Typos on email?
2. Need more background and related work coverage on answering the below questions:
1. What is the history of Synthpop?
2. Any related work on Synthpop already?
3. What is the benefit or advantages Synthpop can bring?
4. What is the new insight this paper bring with Synthpop?
3. The paper uses WDBC(L300) data set to test. What is the background of this data set (how many records? What type of the records? This dataset is used for what?)
4. The synthpop package is effective in the WDBC data set. Please indicate in the paper: Have we done any similar experiments on other medical data sets to see if the package is suitable for larger scales of data sets or in other words, is there any data set the Synthpop cannot handle?
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper mentions a current problem related to medical data sharing and its implication for individuals' privacy and the research community.
Although it focuses on an interesting topic, some contributions require improvements. More than one dataset must be considered in the experiment, selecting a representative set, since the quality of the dataset may influence the quality of Synthpop, and not due to the method. Moreover, the statistical test is not equivalent to the evaluation done in section 4.
In terms of structure, the paper also can be improved. There is a significant missing of references in the overall document, for example, the paper did not mention any references from the analyzed method synthpop (not even the documentation page). It is also recommended to split section 3 into two parts: one to describe Synthpop, in-depth; and another one to describe the entire methodology. The quality of images requires improvements, for example, figure 2 is difficult to read. Finally, the related work describes, in fact, the current background, and it does not describe the related work as similar works published in the literature. Thus, the paper may describe similar works and mention Synthpop alternatives. Also, GAN should be mentioned in the background section, since the abstract mentions it.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Most of my concerns were addressed. Only missing are details on the ML part (specially the GAN topic).
Author Response
Response Letter
We thank the reviewer for reviewing our paper and are grateful for their comments. Please find the response below.
Comments:
Most of my concerns were addressed. Only missing are details on the ML part (specially the GAN topic).
Our response: We thank reviewer for their suggestions, and we have expanded the text related to GANs and have mentioned alternative tools to Synthpop and how they are different. We have also added text related to GANs in the introductions section.
Finally, we thank the reviewers for their valuable and constructive comments to help our paper meet the journal's standards.
