Public Opinions on COVID-19 Vaccines—A Spatiotemporal Perspective on Races and Topics Using a Bayesian-Based Method

Round 1

Reviewer 1 Report

I recommend publication this paper after discussing the following comments:

1. The author should sketch the real data using nonparametric plots to obtain the behaviour of this data. 

2. The authors mentioned that the Bayes estimate was used. Discuss the informative and noninformative approaches. 

3. Abbreviation Section should be reported.

4. Many authors studied the aim of this paper using probability distributions. Discuss this. 

5. The authors proposed their result depending on questionnaire model. Is this true?. If this is true, please write this in a separate Section at the end of the paper.  

6. Some details around the Figures should be reported, specially, Figures 7-12. 

Author Response

1. The author should sketch the real data using nonparametric plots to obtain the behaviour of this data.

Thank you for this comment. We sketched the real data using nonparametric plots by showing the quantity of the tweets in the United States (Figure 5) and the temporal trend of the tweets (Figure 3). We would prefer to keep the original plots unless the “nonparametric plots” in this comment means something else.

 

2. The authors mentioned that the Bayes estimate was used. Discuss the informative and noninformative approaches.

Thank you for this comment. We added the discussion about informative and noninformative approaches in the 4.2 Public Sentiment Score section.

“The Bayesian-based approach integrates prior knowledge and newly observed evidence (i.e., social media information in this research), resulting a reliable estimation on interest variables. Based on the availability of previous information or knowledge, the prior is created with informative or non-informative approaches. When the previous information is available, the prior is modeled as an informative prior, otherwise modeled as a non-informative prior to reflect the balance among the outcomes of interest variables. In this research, the sentiment information in the preceding week is used to model an in-formative prior of PSS for each date”

 

3. Abbreviation Section should be reported.

Thank you very much for the suggestion. An abbreviation section is added at the end of the Paper.

 

4. Many authors studied the aim of this paper using probability distributions. Discuss this

Thank you for the comments. We added how other articles used probability distribution for COVID sentiment analyses in the literature review section 2.1     , and how probability distribution was utilized by us to find the sentiment polarities through the Naïve Bayes method in the 4.2 Public Sentiment Score section.

2.1       Using social media to analyze opinions for public events

Many studies in recent years utilized models and tools which implemented probability distribution to analyze the text sentiment, such as LDA Model and Vader [26,27]. This study utilized Textblob with Naïve Bayes categorization model which also implemented probability distribution concepts to determine the sentiment value of tweets.

4.2 Public Sentiment Score

“Textblob library with Naïve Bayes categorization model was invoked to determine the sentiment polarities. The sentiment polarity probability was determined through the following equation of Naïve Bayes.

P(polarity|Features)=(P(polarity)*P(feature|polarity))⁄(P(features))     (2)

Where P(features) is the probability distribution, which is established from particular features, P(polarity) is the previous possibility of a polarity, and P(feature|polarity) is the previous possibility which particular features is categorized as a polarity.”

5. The authors proposed their result depending on questionnaire model. Is this true?. If this is true, please write this in a separate Section at the end of the paper.

Thank you for your suggestions. Our results do not depend on the questionnaire model. We used nature language processing tool to determine the sentiment of each tweet with uncertainty value.

 

6. Some details around the Figures should be reported, specially, Figures 7-12.

Thank you for this comment; we have added more details around figure 7, 8, 9, and 10, and we included all observations found in figure 11 and 12.

Reviewer 2 Report

I’m thankful to review the paper entitled “Public opinions on COVID-19 vaccines – a spatiotemporal perspective on races and topics using a Bayesian-based method” for Vaccines MDPI Journal.

My decision is Accept after major revision.

 

The paper is interesting but the paper doesn’t conform to the journal’s Author Guidelines. Advanced geographical approaches play a critical role in identifying the locations of disease clustering and vulnerable populations vaccination policies and in implementing effective public health actions.

This study uses data. Data Collection and Confidentiality must be describing more analytic.  Data were collected from the Twitter platform; All data was gathered in compliance with all applicable Twitter API policies and terms of use? The authors must be describing more specific the process of data collection by twitter platform? please declare that the work presented in the article has been carried out in an ethical way. In opposite please add the Ethical approve of the study.

Describe in the methods section the language was used by the participants. English-speaking users in the United States? Spanish or Asian?

The article does not conform to the Guide for Authors for the journal, references are incomplete.

 

Abstract

Abstract should be self-contained and concise, explaining your work as briefly and clearly as possible. Enhance the abstract with more information’s about your results.

Introduction

·        Please insert to introduction-section studies about the vaccine acceptance in pre-endemic period such a: Mfinanga SG, Mnyambwa NP, Minja DT, Ntinginya NE, Ngadaya E, Makani J, Makubi AN. Lancet. 2021 Apr 24;397(10284):1542-1543. doi: 10.1016/S0140-6736(21)00678-4. Epub 2021 Apr 14.

Materials and methods

       Please describe in methods if you exclude retweets

 

Enhance the discussion with the follow references:

1.     Mislove, A., Lehmann, S., Ahn, Y.Y., Onnela, J.P. and Rosenquist, J.N., Understanding the Demographics of Twitter Users. ICWSM, 11, p.5th., 2011.

2.     Hecht, B., and Stephens, M. 2014. A tale of cities: Urban biases in volunteered geographic information. ICWSM’14, 197–205.

3.     Scaling laws in geo-located Twitter data. Arthur R, Williams HTP. PLoS One. 2019 Jul 24;14(7):e0218454. doi: 10.1371/journal.pone.0218454.

4.     The human geography of Twitter: Quantifying regional identity and inter-region communication in England and Wales. Arthur R, Williams HTP. PLoS One. 2019 Apr 15;14(4):e0214466. doi: 10.1371/journal.pone.0214466. e Collection 2019.

 

      Line 353 Result and Discussion: reform us follow: Results and Discussion in separate paragraph according to Journal suggestions.

      Line 420-422: “This is likely because there were frequent COVID policy changes and more COVID news reports compared to the other states”. Please improve this approach with specific references.

      Line 445-449 according to authors “In the temporal trend of nation PES, there are 4 spikes in Figure 3. The third sub-period is between the date of the very first dose and the date of reaching 100 million vaccines administered. This sub-period represents when the public started the vaccination process. The last sub-period is after the date of reaching 100 million vaccines administered. This sub-period is a milestone that represents the time many people had been vaccinated”. Could authors explain the peak between April and May and after the milestone of 100 million vaccines administered?

Limitations: please add limitations for the study in separate paragraph. Rus and Cameron [Rus HM, Cameron LD. Health communication in social media: message features predicting user engagement on diabetes-related Facebook pages. Ann Behav Med. 2016 Oct;50(5):678–89. doi: 10.1007/s12160-016-9793-9] showed an older study that messages with images had higher rates of liking and sharing relative to messages without images. Furthermore, the formats and demographics of social media sites are constantly changing and have evolved since the study was conducted. Some social platforms such a Twitter users tend to be younger and more educated compared to the general population. In the tweets may not be truthful and may introduce elements of bias into the data. Mislove et. al. found “Twitter users are more likely to live within populous counties than would be expected from the Census data, and that sparsely populated regions of the US are significantly underrepresented”. Hecht et. al. “find that urban users are over-represented in geo-located social media, and also provide more information than rural users”. The authors didn’t report any limitation for this study.

References

Follow the instructions of the journal, all references must be reform:

·        References must be numbered in order of appearance in the text (including table captions and figure legends) and listed individually at the end of the manuscript. We recommend preparing the references with a bibliography software package, such as EndNote, Reference Manager or Zotero to avoid typing mistakes and duplicated references. We encourage citations to data, computer code and other citable research material. If available online, you may use reference style 9. below.

·        Citations and References in Supplementary files are permitted provided that they also appear in the main text and in the reference list.

 

In the text, reference numbers should be placed in square brackets [ ], and placed before the punctuation; for example [1], [1–3] or [1,3]. For embedded citations in the text with pagination, use both parentheses and brackets to indicate the reference number and page numbers; for example [5] (p. 10). or [6] (pp. 101–105).

Author Response

1. The paper is interesting but the paper doesn’t conform to the journal’s Author Guidelines. Advanced geographical approaches play a critical role in identifying the locations of disease clustering and vulnerable populations vaccination policies and in implementing effective public health actions.

Thank you for this comment; We just modified the paper to conform to the journal’s Author Guidelines.

 

2. This study uses data. Data Collection and Confidentiality must be describing more analytic. Data were collected from the Twitter platform; All data was gathered in compliance with all applicable Twitter API policies and terms of use? The authors must be describing more specific the process of data collection by twitter platform? please declare that the work presented in the article has been carried out in an ethical way. In opposite please add the Ethical approve of the study.

Thank you. We added the process of how we collect the data in the following paragraphs.

The Harvard Center for Geographic Analysis (CGA) maintains the Geotweet Archive, a global record of tweets spanning time, geography, and language. The primary purpose of the Archive is to make a comprehensive collection of geo-located tweets available to the academic research community. The archive extends from 2010 to the present and is updated daily and the number of tweets in the collection totals approximately 10 billion. The data is collected using Twitter’s Streaming API following Twitter’s Developer Agreement and Policy [44].

The Geotweet Archive consists of tweets which carry two types of geospatial signature: 1) GPS-based longitude/latitude generated by the originating device 2) Place-name-centroid-based longitude/latitude from the bounding box provided by Twitter, based on the user-defined place designation [44].

Any tweet which carries one or both signatures is included in the Archive. Approximately 1-2% of all tweets contain such geographic coordinates. The current version of the Archive is Version 2.0. The original Version 1.0 archive began in 2012 as part of a project to develop a GPU-powered spatial database called GEOPS. Version 2.0 of the archive represents the results of a merge between the CGA archive, and an archive developed by the Department of Geoinformatics at the University of Salzburg in Austria, as well as several other archives [44].

For the purposes of ethical approval, this study removed specific twitter identity for protecting privacy of users after generating statistical results. Only statistical results are presented in this article.

3. Describe in the methods section the language was used by the participants. English-speaking users in the United States? Spanish or Asian?

Thank you for suggestions. We added the language that was used in this research. We think adding this part into the data section might be better than the method section, so we added the following description in the 3.1 Social media data section.

The data excluded retweets and were all English-speaking tweets in the contiguous United States which covers 48 adjoining states and the district of Columbia.

4. The article does not conform to the Guide for Authors for the journal, references are incomplete.

Thank you for your suggestions. We double checked the reference section

5. Abstract should be self-contained and concise, explaining your work as briefly and clearly as possible. Enhance the abstract with more information’s about your results.

Thank you for your suggestions. We added more information about our results in our abstract

6. Please insert to introduction-section studies about the vaccine acceptance in pre-endemic period such a: Mfinanga SG, Mnyambwa NP, Minja DT, Ntinginya NE, Ngadaya E, Makani J, Makubi AN. Lancet. 2021 Apr 24;397(10284):1542-1543. doi: 10.1016/S0140-6736(21)00678-4. Epub 2021 Apr 14.

Thank you for your suggestions. We added vaccine acceptance discussion in pre-pandemic period but used a different article as reference: Yaqub, O., Castle-Clarke, S., Sevdalis, N. and Chataway, J., 2014. Attitudes to vaccination: a critical review. Social science & medicine, 112, pp.1-11. And we also cited the reference recommended to demonstrate the concerns on vaccines from Tanzania’s publics.

7. Please describe in methods if you exclude retweets

Thank you for your suggestions. We added descriptions in section 3.1 Social media data that we excluded retweets.

8. Enhance the discussion with the follow references

Thank you for your suggestions. We included the Arthur’s article in the literature review section, and included Mislove and Hecht’s article in the limitation section

9. Result and Discussion: reform us follow: Results and Discussion in separate paragraph according to Journal suggestions.

Thank you for your suggestions. Our paper tried to follow the journal suggestion on the section formatting in its template. However, our contents need more sections than how the journal suggested. In addition, the results of our experiments are presented by numbers and figures. It is hard to separate the results and discussion into different sections in this article. Therefore, we would prefer to keep the section format.

10. “This is likely because there were frequent COVID policy changes and more COVID news reports compared to the other states”. Please improve this approach with specific references.

Thank you for your suggestions. We added the reference needed for this statement

11. according to authors “In the temporal trend of nation PES, there are 4 spikes in Figure 3. The third sub-period is between the date of the very first dose and the date of reaching 100 million vaccines administered. This sub-period represents when the public started the vaccination process. The last sub-period is after the date of reaching 100 million vaccines administered. This sub-period is a milestone that represents the time many people had been vaccinated”. Could authors explain the peak between April and May and after the milestone of 100 million vaccines administered?

Thank you for your comments, we added the description of the spike between April and May. It is caused by the news that Johnson and Johnson decided to pause its vaccination all clinic trials.

12. Limitations: please add limitations for the study in separate paragraph.

Thank you for your suggestions. We previously discussed the limitations of this research in the last paragraph of the conclusion, but we added more limitation based on the reference you suggested and separate the limitations into a separate paragraph (the second last paragraph)

13. References

Thank you for your suggestions. The reference order is adjusted.

Round 2

Reviewer 2 Report

Accept in the present form.

Thank you for the revised manuscript.