iCite is a powerful web application that provides a panel of bibliometric information for journal publications within a defined analysis group (where an analysis group can consist of a single article or a very large group of articles). The data produced by iCite can be downloaded as a customized report from the dashboard and could be used to understand the influence of articles within an analysis group. An example application for iCite might be to compare how the influence of a portfolio of articles compares to the remaining articles that come out of grants funded by the NIH.
The following data are produced using iCite:
The Relative Citation Ratio is a new metric developed within the Office of Portfolio Analysis (OPA) that represents a citation-based measure of scientific influence of one or more articles. It is calculated as the cites/year of each paper, normalized to the citations per year received by NIH-funded papers in the same field and year. This benchmarking process, performed with quantile regression, ensures that a paper with an RCR of 1.0 has received the same number of cites/year as the median NIH-funded paper in its field, while a paper with an RCR of 2.0 has received twice as many cites/year as the median NIH-funded paper in its field. The displayed values are the average and standard deviation of the papers in the group along with the median.
iCite is limited to analyzing only articles that appear in PubMed; users upload the PubMed IDs for articles within the analysis group of interest (articles identified from either SPIRES or PubMed searches). The citation database powering iCite is generated from a number of sources, including CrossRef, MedLine, PubMed Central, and Entrez. At present, only PubMed citations are included, so citations appearing from journals outside PubMed are not counted. iCite allows for the analysis of just one group of articles or for the comparison of two separate groups as appropriate.
Once the data have been processed, two summaries appear on screen: a table with summary data, and graphs with a year-by-year breakdown.
Metrics describing the set of selected papers that were uploaded are displayed in this table. The metrics given in the table are Total Pubs, Pubs/Year, Cites/Year, RCR and Weighted RCR. Total Pubs, Pubs/Year and Weighted RCR are given as single values, but for Cites/Year and RCR, the maximum, mean, Standard Error of the Mean (SEM), and median are all shown. Users can click on the column labels to see detailed descriptions of how these metrics are calculated. These are summary statistics for the whole group, but yearly breakdowns are shown for Total Pubs, Mean RCR, and Weighted RCR in the graphs below.
Below are the descriptions that are shown when users click on the column labels:
These histograms show:
There are some options for selecting and deselecting subsets of the data based on publication year or type. These options are available if there are fewer than 1000 publications.
The year ranges give the option of selecting only documents in certain time windows. The "Exclude non-articles" check box deselects papers flagged as non-articles. Values from these de-selected papers will not be included in the calculations below. After making changes here, the summary information and graphs are automatically updated.
To be counted as an 'article', a document must have one of the below PubMed "Publication Type" tags indicating a possible research article, and none of the below "Publication Type" tags indicating derivative or non-research documents:Must have at least one to be counted as 'article':
Based on manual curation of a random sample of articles published in 2013, this classification strategy yields 92% accuracy. 8% of 'article' assignments on average were incorrect, while 5% of 'non-article' assignments were incorrect. Most errors tend to occur in journals that are not indexed in MedLine, suggesting that persons whose publication records are concentrated in non-MedLine journals may have disproportionately high error rates in this classification schema. We are working to address this issue.
Note that in some cases, PMIDs cannot be matched to citation records. In this case, a warning appears:
Clicking "View details" displays the list of missing articles and common reasons why the PMIDs were not matched:
Most often, this occurs because articles are outside the year range for available citation data. We currently have citation data for papers from 1995-present, but note that papers published in the most recent year (some of which are only a few months old) have not had enough time to accrue a meaningful citation count. For this reason, only articles that are at least a year old are available for calculating RCR. Recent papers may also show stronger seasonality patterns and discrepancies because of online publication preceding print publication. Papers that are published early in the calendar year or with long online-only publication periods will have a longer period in which to accrue citations.
If your dataset is larger than 1000 papers, you are limited to the basic selection options described above. However, if your dataset has fewer than 1000 papers, there are further customization options. You can see the title, authors, article type, journal, and RCR of each paper:
Users can select and deselect individual papers with the checkbox on the left (for example, deselecting a person's middle-author papers). In addition, whole years can be selected or deselected. "Select/deselect all" can be used to reset the individual selections.
iCite users can download the article-level data underlying the summary statistics. Use the "Download selected results" button to download the selected results as a spreadsheet.
Clicking this button will download the selected (but not the deselected) articles as a spreadsheet with additional bibliometric information. This will appear in your web browser's default download folder.
To compare two or more groups in iCite, users must upload a two-column spreadsheet: one column with PMIDs and one column with group labels. For this example, publications were selected that were linked to NIH grants falling into one of two Program Class Codes (a neuro-development set and a neuro-endocrinology set):
As with an analysis of a single group, a table appears toward the top of the page with bibliometric information displayed:
For multiple groups, summary statistics are calculated for each group separately. If there are multiple groups, toggle between graphs for each group by selecting the eye icon for your chosen group.
The procedure for downloading data is the same for multiple groups as it is for single analysis groups. However, the downloaded file will contain a column with the group labels so that the publications can be analyzed outside of iCite (for example, with Pivot Tables in Excel).
We seek to support all modern browsers and platforms. If you experience a problem using the current version of any of the following, please let us know:
Send an email to iCite@mail.nih.gov.