Original link: https://yufree.cn/cn/2023/07/12/c-index/
After the impact factor of journals came out this year, there was a dead silence on the Internet, as if this hadn’t happened. The reason is very intuitive. This year, the impact factor of 80% of academic journals has dropped. Last year, the impact factor generally rose, and almost all journals announced the good news. Although everyone likes to read outrageous news this year, things that are closely related to themselves still don’t like this kind of downhill news.
But I don’t think it matters. The impact factor is a method of reviewing journals. There is nothing wrong with the excitement of academic publishing companies. But when it comes to individuals, they should not use the method of reviewing journals to judge people. Most people who say how many high-scoring articles they have published quote The number should not reach the average value of the journal. If it is enough, he will say how many citations his article has received in a year. However, there is indeed a reason for the general drop in the impact factor, and it is most likely related to the new crown.
When the impact factor came out last year, the number of citations in 2020-2021 was calculated. I don’t know what others are doing at this time, but I could only write papers at home. Another reason lies in the calculation method of the impact factor, that is, the number of citations divided by the number of articles. There is no way to manipulate the number of articles, but the number of citations can be brushed. The common method is to write non-peer-reviewed editorials or Looking forward to introducing work, this trick has been used by major journals such as CNS, and now many small journals have also learned this routine. For example, if I publish a paper, I will find someone to write an introduction and citation in the current journal. Comprehensive journals usually have such a column introducing research progress, so your article will most likely be cited in the current issue when it is published in the top journal.
Originally, what this calculation method can actually contribute is self-citation, and the number is limited. However, in the short term after the COVID-19 pandemic, tens of thousands of articles or preprints related to the COVID-19 pandemic have been produced, and major countries have also started research and money-spending models to promote research and development. Relevant research has triggered a large number of media follow-up and reports, and many of the media are comprehensive journals. Assuming that a COVID-19 paper cites thirty or forty papers, and after the paper is published, the citations are only three or four (this is the average level, and it is not black if it is not blown), then the essence of this paper is to inject thirty or forty citations into academic journals. Most of them flow to mid-to-high-end journals. Under this trend, the impact factors of mid-to-high-end journals must be getting higher and higher. Specifically, last year, all the journals of basic biomedical research related to the new crown have generally increased by a large margin.
Without evidence, let me talk about the data. In 2022, when the impact factor is soaring, the two-year data from 2020 to 2021 will be used. In 2021, there will be 142 covid-related papers in the journal Science, and there will be 175 in 2020. If we only look at the citations of the year, we will find that there were 7,673 citations in 2020, and each article was cited about 44 times. In 2021, there were 4,006 citations, and each article was cited about 28 times. What about all papers? “Science” published 2,656 articles in 2020, with a total of 19,788 citations in that year, 2,636 in 2021, with a total of 12,943 citations in that year, and 2,496 in 2022, with a total of 10,343 citations in that year, compared to 2,729 in 2019 before the epidemic. The total number of citations that year was 7217. What is the concept? That is to say, the number of citations of all papers on “Science” in 2019 is about the same as the number of citations of covid-related papers in 2020. So there is no doubt that the new crown has brought a wave of large traffic to scientific research papers, which is the main reason for the expansion of the impact factor in the past two years.
So what happened this year? The main reason is that research attention has declined, and the number of related research projects has returned to normal levels. Last year (2022) “Science” published 2,496 papers, 85 of which were on Covid-related topics, and there were 1,076 citations that year, and each paper was only cited about 13 times. But unfortunately, the publication of many follow-up papers is lagging behind. The number of citations of these new crown papers cannot be increased after they were published last year, and the publication of these articles occupies the number of regular papers. The consequence is naturally a wave of decline. Comprehensive and biomedical categories fell the worst. This year’s statistics also include 2021, the year of citation inflation, but by the end of the year at most, these will return to 2019 levels. But because of the halo effect of the number of citations I just mentioned, the overall mid-to-high-end journals will still attract more citations. The dilution effect of new journals uploaded on the Internet is not a bit worse than the impact of the new crown. Good research publications in any medium-to-high-end journals will be cited for a long time. More papers will only make these high-quality content cited more. At the journal level The effect of is meaningless to researchers.
Having said that, I will go back to the original point. How to evaluate the paper? The trouble now is that no matter what indicator you use, you may be brushed. Whether it is the number of citations or the number of papers currently being investigated, they are all one-way indicators, that is, the higher the better, a monotonic function will be brushed no matter what. The most common method is to send a bunch of low-end papers to cite references, or write a review to cheat citations. But if you give an indicator, the higher the value, the higher the difficulty, it will be much better. For example, the H index is not easy to brush. But the H index cannot evaluate a single paper, so I thought about designing a more general C index.
The calculation method of the C index is very simple, that is, the number of times the article is cited (based on the evaluation time) divided by the number of cited papers. If it exceeds 1, it means that the quality of this paper is relatively high. After all, the number of people citing this paper is higher than this There are even more citations in the article, which expands the influence of the discipline. Among the papers published by each person, the number of papers with a ratio of more than 1 is accumulated as the C index of this person, and journals can also be counted in this way.
The threshold for this C-index is still very high. An academic paper supported by an experiment can have 30 to 40 citations, and being cited 30 to 40 times is a high-quality paper for any journal. Moreover, the biggest advantage of the C index is that it can shield the influence of review articles. Nowadays, a review article often has hundreds of citations, but a review that can get hundreds of citations after publication must be a milestone. Most reviews are not enough. to this level. In addition, the basic assumption of the C index is that if a piece of research is pioneering, then its cited work must be less than the cited work. If a research paper cites hundreds of papers, then the biggest possibility is that the experimental evidence is not enough, and you can only use other people’s work as evidence. If you use this index, you can also shield the unreasonable demands of many reviewers to add citations to increase your own citations, because the added citations will automatically increase the threshold of your own article’s C index. Moreover, this indicator is naturally anti-spoofing. You can only add one if you do original work and get the approval of reviewers and citers. If you simply do permutation and combination work, you will definitely have more self-citations, and the more you self-cite work , the lower the C index. Using the C-index evaluation can make the author more cautious in citing instead of citing all over the place, and pass by pleasing the reviewers. This kind of paper is also the batch that lowers the average citation for the journal. For individuals or journals, the higher the C index, the more original work, and it is more worthy of reading by those who pay attention to the frontier. However, the review with C index can focus on studying, and it will definitely not make mistakes, and it is also suitable for teaching people to read.
I guess that the C index of most researchers may not reach double digits. After all, most people do follow-up research rather than pioneering research, and the number of citations for follow-up research is naturally higher than that of pioneering research. If you agree that I will control the number of citations in the future, it depends on whether the reviewers are happy, because if the number of follow-up research citations is small, the reviewers will think that the background knowledge is not enough or the quality is not enough. The only problem with the C index is how to calculate the number of citations. This is not difficult. Each discipline has its own half-life of citations. Just look at the C index that can only be used for the half-life of citations after the article is published. Some disciplines have a shorter half-life, and the C index is more difficult to earn, while those with a long half-life naturally require your work to have long-term value. Therefore, considering the C index of the half-life in this way should be comparable across disciplines. Although I think it is not meaningful, many times evaluating scientific research results is a matter of comparing apples to oranges.
At this point, we can look back at the C index of papers on Science. Here we set the half-life as two years, which is consistent with the impact factor calculation. I will not check individual references, but use the average value. In 2019, 2,729 papers cited a total of 112,842 papers, each citing an average of 41 reference papers. How many citations will exceed the number of article citations in 2020-2021? About 550-600 articles, that is, the C index is about 550-600. The average citation in 2020 is 45. It can be seen that there is not much impact from the new crown. The C index in 2021-2022 is also about 550-600. In other words, this index is naturally immune to hot spots. Regardless of whether there is a major trend or not, the number of cited papers and the number of referenced papers by scholars as a whole are relatively stable. The influence is as great as “Science”, and the C index does not change much every year. About 20% of the papers are truly original.
As a comparison, let’s look at ES&T, which is the top journal in environmental science. In 2019, 1,548 articles were published, 43,981 articles were referenced, and the average references were 28.4 articles. Then the C index is about 150-200, and the original proportion is about 13%. In 2020, 1,691 articles were published, 35,276 articles were referenced, the average number of references was 21, the C index was between 250-300, and the proportion of original works reached about 20%. That is to say, there is not a big gap between the original high-quality papers in the top publications of sub-disciplines and the original high-quality papers in comprehensive journals. Mainstream comprehensive journals maintain 20% of original high-quality articles on a daily basis. When major scientific problems arise in sub-disciplines It is possible to achieve this ratio, but it may be a little worse in normal times. At the same time, it should also be noted that about 80% of the top journal papers actually have limited influence. It is likely that they are flooded by related households or lucky, and they cannot reach the average level of journals or shake the foundation of the discipline.
Personally speaking, people who can get a double-digit C index are probably already professors, which means that they have published more than 10 original results, or the H index is at least twenty or thirty. If you only look at one work, most professors have already established a research group with less than two digits in an article, and as a corresponding author, it is hard to say that the originality of the student, postdoctoral fellow or the professor himself, but the C index evaluates a research group There is no problem, at least a C index greater than 1 can be counted as an original research group.
I’m not worried that the C index is not easy to calculate, but as a wild evaluation index, at this point in time, no one should “optimize” and it is not easy to optimize. You can count it for fun. I wouldn’t be surprised if it becomes a recognized indicator one day. After all, scientific research must be based on pioneering rather than follow-up work.
This article is transferred from: https://yufree.cn/cn/2023/07/12/c-index/
This site is only for collection, and the copyright belongs to the original author.