SeminarSeminar

SUB MENU

Seminar Questions and Answers for Instructors

The 24th JANS Seminar
Let's get to know the concept of variable selection." Causal Reasoning Recommendations."

Question 1.
You say that RCTs can adjust for unmeasured factors, but is this possible in RCTs? Do you think it is possible to adjust by randomization or, for example, by prospectively obtaining information on the factor you want to measure?
Your explanation of the propensity score was very clear and informative.

Answer:Thank you for your question.
The word "adjusted" has the implication of "statistically adjusting for the effects of confounders. The exact phrase "RCTs can adjust for unmeasured confounders" would be "Randomization can be assumed to balance the assignment of unknown and unmeasured confounders to each group, so there is no need to consider their effects.
Specifically, in the case of an RCT, the survey items shown in the subject background table will have a similar distribution in the two groups (they may be biased by chance in probability). Also, unknown and unmeasured factors that are not shown in the table (not in the survey items) are assumed to have similar distributions in the two groups as well. Therefore, no "statistical adjustment" is required in an RCT.

Question 2
I was wondering why it was not given as a method in this case, although it was mentioned that there is a way to put in a propensity score as a covariate. Does that mean that method is not very appropriate?

Answer:Thank you for your question.
As you point out, some studies do throw in propensity scores as covariates.
Since the two methods compare different effects (with or without subjects removed by matching), the results may differ. There are some papers that report both results together, but there do not seem to be any recent studies that adjust for this as a covariate. We believe this is because the propensity score matching and inverse probability weighting methods make it clear (to some extent) what effects they are looking at, but it becomes difficult to understand when adjusted as a covariate.

Question 3
It was very easy to understand, but I do not have a perfect understanding of variable selection. I have no idea about confluence bias and so on. I would like to have a consultation with a statistician like Dr. Shuntaro Sato when I actually analyze the data, as is the case in other countries.

Answer:We will share your requests with the members of the Research and Academic Advancement Committee.

The 20th JANS Seminar
Advances in Open Science and the Future of Nursing: Bringing Open Data to Nursing Research

Question 1.
It seems to me that there are ethical issues involved in archiving and making available historical data. The research participants have provided us with their data with an understanding of the significance of the study. I think there are ethical issues in using that data for secondary analysis unless it is explained at the time that this data will also be used for secondary analysis.

Answer:In this regard, archiving (long-term preservation) and secondary use of data should first be considered separately. Disposal of publicly funded research data is not at all recommended by the standards of modern archiving and open science.

As you say, the issue of secondary use remains. It goes without saying that, ideally, permission for publication and secondary use should be obtained along with the survey responses. If the secondary data is not the data itself, but secondary data that has been processed in various ways, the possibility of publication and secondary use can be opened up. The next question is the degree of processing, and in order of decreasing difficulty, the following will be considered: release as aggregate data, release as anonymous data, and release as raw data. One benchmark is the availability of anonymized microdata from official statistics, even though permission for secondary use is not obtained at any given time.

However, this also depends on the context in which the survey was conducted. In my lecture, I assumed that the survey was conducted on a nationwide scale, with a random sampling of several thousand subjects, and that the distance between the survey subject and the survey participants was very great. If the data were directly collected at a specific hospital, and the number of subjects were relatively small, anonymization would be practically difficult, and disclosure and secondary use would be even more difficult.

In the end, each field has its own approaches and norms, and it will be important to develop guidelines for data disclosure that are appropriate for each field. While it is a matter of course to protect the personal information and privacy of survey subjects, it will become increasingly important to consider how to share data, which are public goods, with the rest of society.

Question 2
What are your thoughts on the openness of qualitative research data such as interviews and fieldwork (ethnography)? Are there currently any databases, etc. being built for qualitative research data as well?

Answer:First of all, the openness of qualitative data (*1) is sure to advance in the future in the context of the open science (*2) trend surrounding the academic world. In fact, we have seen an increase in the number of academic publications on secondary analysis of qualitative data in the past few years (*3).
As for databases/archives of qualitative data at present, there is no organization in Japan that stores qualitative data collected through academic research or provides them for secondary use. However, in recent years, there have been activities to open up qualitative data by limiting the fields, themes, and types of data under different management systems and scope of activities from general data archives, such as DIPExJAPAN (*4), which handles "data on narratives of health and illness".

Overseas, starting with the Qualitative Data Archival Resource Centre (a.k.a. Qualidata) established in the United Kingdom in 1994, qualitative data archiving has been conducted in Europe since the beginning of the 2000s, including Finnish Data In the 2000s, qualitative data archiving began to take place in Europe, including Finnish Data Service in Finland and Swiss Data Service in Switzerland. In Asia, the Korea Social Science Data Archive in South Korea has been handling qualitative data since 2010 (*5).

(*1) Archiving and sharing qualitative data with a view to secondary use.
(*2) Efforts to make research data and results widely accessible to experts and non-specialists, mainly through the use of ICT.
(*3) For example, Beck (2019) Secondary qualitative data analysis in the health and social sciences.
(*4)https://www.dipex-j.org/
(*5) In other countries, there are not only cases where new qualitative data archives are established, but also many cases where existing data archives have started to handle qualitative data.
Incidentally, Qualidata merged with the UK Data Archive in 2001.

Question 3
Disclosure of qualitative data often involves content that is closely related to the personal background of the research subjects. We would like to know what points need to be considered in the procedures for processing the data into data that can be disclosed.

Answer:I think it is very important to point out that the processing of qualitative data for secondary use (anonymization, processing of images and sounds, etc.) varies depending on the type of data, research theme, when the data was collected, what kind of consent was obtained, etc., so it is difficult to give a generalized answer to your question. In fact, it is difficult to give a generalized answer. In fact, the processing of qualitative data itself constitutes a research theme (*1).

With this preamble, and with interview data in mind, I will answer in a simplified form: we will be balancing the value of the data and the privacy of the participants while deciding "to what extent to anonymize the information in the transcripts". In doing so, information to be particularly careful about includes, for example, names, locations, religious backgrounds, political beliefs, occupations, and family relationships.

(*1) For example, Saunders, Kitzinger, & Kitzinger (2015) Anonymising interview data: challenges and compromises in practice.

Question 4
In order to ensure the reliability of qualitative data research results, please provide any examples of how various researchers have utilized the same data, analyzed them separately, compared results, and derived more accurate results.

Answer:First of all, we are not aware of any cases regarding your question. Then, I would like to add some additional information. When the objective is to ensure or improve trustworthiness (=trustworthiness, rigorousness, validity), the following two types of qualitative data may not be considered in the first place. For example, if the purpose is to ensure or improve reliability (=trustworthiness, rigorousness, validity), it may not be considered a secondary analysis of qualitative data in the first place. For example, the method of comparing the results of analyses by different researchers within the same research team with the same research objectives is known as triangulation, which is not considered a secondary analysis because it has the "same research objectives. On the other hand, if the purpose is to "verify" the reliability of past results, it may be considered as a pattern of secondary analysis (*2). (*2) However, there are various views among researchers as to the extent to which the idea of "validating" the results of past qualitative research is appropriate in itself (*3).

(*1) For example, Heaton (2004) Reworking qualitative data takes the view that qualitative meta-analysis is not a secondary analysis for similar reasons.
(*2) Thorne (1994) Secondary analysis in qualitative research: issues and (*2) Thorne (1994) Secondary analysis in qualitative research: issues and implications calls such secondary analysis "cross-validation", while Corti & Thompson (2004) Secondary analysis of archived Corti & Thompson (2004) Secondary analysis of archived data calls it "verification".
(*3) For example, it is mentioned in Corti & Thompson (2004) mentioned above.

17th JANS Seminar
Social Implementation of Nursing Research: Trends in Implementation Research and Data Science

Question 1.
Which vendor would you recommend for an electronic medical record that would be effective for data extraction, analysis, and other research efforts?

Answer:If DWH is an optional feature, the facility may not be able to implement it due to budget constraints. The availability of staff with the skills to perform the extraction may also play a role in whether or not the data can be extracted. Other options include contracting with a vendor to perform the data extraction, but again, budget may be a factor.

Question 2
I am not sure whether to choose Python or R for learning analytical skills. Therefore, we would appreciate it if you could tell us why Python is your first recommendation and R is your second recommendation.

Answer:I made Python my first recommendation because of the amount of information available on the Internet and its potential for future development. In my opinion, Python is better if you mainly want to do machine learning (e.g., artificial intelligence development), and R is better if you mainly want to do data analysis research. I use both depending on the application.