Hi,

I would say, Co-Citations analysis is better for this type of research, since it clusters the main contributors to the field. Keyword analysis is not that focused on the field, it is scattered, you could have the same keywords for two different science field even if they are not close. For instance the keyword "education" could appear in physics, and in sociology, and in computer science, with different meaning in each field. so, in my opinion, keywords do not present an accurate picture of a theme or a research field.

Dear Sardar Wasi Uddin Al Ahmed

It's a really interesting question. Keyword co-occurrence analysis should, theoretically, provide a better understanding of the research area; however, the ambiguity often leads to degraded results. Co-citation analysis should provide a better result (bearing in mind that keywords provided in many articles may be ambiguous and generic). This is something I agree with Avishag Gordon on. On a side note, Co-citation analysis, unfortunately, is also prone to bias. A network or heatmap, for example, is very independent of a particular thematic area, but they can influence such an analysis in terms of overestimation. A second reason for using Keyword co-occurrences is that they are direct indicators of the themes. In addition, a good sample size is more likely to eliminate most of the outliers. Having said that, I would like to review both analyses before deciding which one to use and it greatly depends on the dataset I am working on. As an example, I may prefer Keyword co-occurrence analysis when I am working on data from PubMed where I have the opportunity to incorporate Mesh keywords that are curated and scientific. I would go with a co-citation analysis if, however, I knew that the dataset I was working on may have overlapping keywords. Thanks

I would agree with Avishag Gordon basically.

But you should also take into account the timespan of your analysis. If it is broad it may happen that one cited research item would be replaced in the latter articles with a corresponding newer one (showing the same problem/solution but in more fresh light). In this case, the co-citation method will not indicate these articles as co-cited. Then it is worth considering supplementing co-citation with the keyword co-occurrence method. Also, mind that quite frequently metadata describing scientific works include not only keywords provided by authors but also more 'canonical' ones provided by editors/publishers.

Hi,

It is true that new material in a sample would not be co cited, but most researchers are choosing their sample from earlier periods, and if they have doubts about co-citations as reflecting the boundary lines of a subject they can repeat the test for several periods. The keywords are provided by the authors, by the publishers and/or editors, but also by the database producers, so they reflect more than one view, and in this way it is difficult to draw the boundary lines of a research field or a subject area.

It would be prudent to use both in order to capitalize the strength of each.

I believe keyword co-occurence analysis has an edge over the other

Keyword co-occurrence analysis. Prior to the analysis, the data should be screened and cleaned.

"Keyword co-occurrence analysis" will be better as per my knowledge as you are analyzing thematically. citation may or may not depends on theme or topic.

Satyabrata Sahoo I agree with you!

Wouldn't it be worth adding the bibliographic coupling method to this discussion?

Sardar Wasi Uddin Al Ahmed, Avishag Gordon what would you say?