Application Of Causal Inference In Scientometrics

As correlation analysis has become a basic research method in Scientometrics,scholars have found that correlation does not represent causality,and the conclusions drawn from correlation analysis are of limited use in solving specific problems or making policy recommendations.Therefore,this paper introduces in detail how causal inference is applied to Scientometrics.Causal inference,as a new research method,not only allows researchers to have a cognition of the evolution between events and the internal logical relationship process,but also can further provide intervention guidance and decision analysis for related research in the field of Scientometrics.On the basis of combing and summarizing the research status and basic knowledge of causal inference,this paper introduces the basic principles of three common causal analysis methods in the process of capturing causal effects.At the same time,taking the three methods: Granger causality inference,propensity score matching and differential method as examples,the data requirements,mathematical derivation and concrete implementation steps of obtaining the "net" effect between things are elaborated in detail.In addition,this paper introduces three methods into the hot research field of scientometrics by means of empirical analysis:1.In the Lancet journal as the research object,this paper applies Granger causality to the relationship between downloads and citations of scientific publications,and from the high and low cited papers,cited patterns and different types of paper,further explore the different categories paper downloads and citations the interaction relationship between the number of differences between the causal effect.The results show that there is an obvious causal relationship between downloads and citations,and it tends to be a two-way causal relationship.The causal relationship between monthly cumulative downloads and monthly cumulative citations is more significant.Compared with papers with low citation frequency,the correlation between the number of cited papers and the number of cited papers with high citations is particularly obvious and tends to be a two-way causal relationship.At the same time,the causal relationship between monthly downloads and monthly citations is affected by different citation patterns.The location of the peak number of citations in the life cycle of citation often determines the direction of the interaction between monthly downloads and monthly citations.In addition,it is also found that the causal relationship between the monthly cumulative downloads and the monthly cumulative citations is related to the growth rate of the two.In the same period of time,the indicator with a higher growth rate tends to influence the Granger cause of the indicator with a lower growth rate.2.Taking the Nobel Prize in Physics as an example,this paper applies the propensity score matching method to the citation effect study.The results show that winning the Nobel Prize in Physics has a significant positive effect on the citations of the laureates' scientific publications,that is,after an article is published,the citations of the articles before the award will be significantly increased if the author wins the Nobel Prize.Winning the Nobel Prize in Physics can promote the citation times of scientific publications in most research fields of physics,which greatly promotes the dissemination of the physics knowledge achievements and the discovery of new knowledge of these laureates.3.Using the National Institutes of Health(NIH)medical funding as an example,the difference in difference is used to capture the causal effect of receiving NIH funding on the number of citations of scientific publications.The main findings are as follows: Overall,NIH funding has a significant causal effect on the number of citation times of medical researchers' articles;Different types of NIH funds also have different causal effects on the citation times of articles.Among them,funding projects used for students' scientific research training have the greatest promotion on the citation times of articles of sponsors.Researchers from countries with higher levels of medical services have a more significant increase in citation times after receiving NIH funding.Writing in English,which is widely disseminate,can also improve the dissemination of knowledge.Scientists at medical research centers,which are more focused on medical research,saw the highest increase in citations after receiving NIH funding.