Modelling count response variables in informetric studies: Comparison among count, linear, and lognormal regression models

Ajiferuke, Isola; Famoye, Felix

doi:10.1016/j.joi.2015.05.001

Cited by 42 publications

(24 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Regression analysis was conducted for citation counts (for 6-month, 12-month, and 24-month citation windows) and altmetric counts (for tweets, blogs, mainstream media articles, Wikipedia mentions, and Mendeley reads) using a negative binomial regression model, with the full set of explanatory variables as described above. A negative binomial regression model is more suitable for over-dispersed count data (as is the case with citation and altmetric count data) than a linear regression model (Ajiferuke and Famoye, 2015). Regression was conducted using the R package MASS (Venables and Ripley, 2002).…”

Section: Methodsmentioning

confidence: 99%

The effect of bioRxiv preprints on citations and altmetrics

Fraser

Momeni

Mayr

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

Section: Methodsmentioning

confidence: 99%

The effect of bioRxiv preprints on citations and altmetrics

Fraser

Momeni

Mayr

et al. 2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Moreover, the random slopes better accommodate potential differences in bivariate relationship strength between predictor and dependent variables, rather than forcing them to a common average among journals. WS citation rates were modeled as a random variable of a negative binomial distribution [26]. Predictor variables were considered to have statistically significant effects if the 95% confidence intervals associated with their parameter estimates did not bound zero; this approach is analogous to a frequentist significance test at α = 0.05.…”

Section: Methodsmentioning

confidence: 99%

Twitter Predicts Citation Rates of Ecological Research

et al. 2016

View full text Add to dashboard Cite

The relationship between traditional metrics of research impact (e.g., number of citations) and alternative metrics (altmetrics) such as Twitter activity are of great interest, but remain imprecisely quantified. We used generalized linear mixed modeling to estimate the relative effects of Twitter activity, journal impact factor, and time since publication on Web of Science citation rates of 1,599 primary research articles from 20 ecology journals published from 2012–2014. We found a strong positive relationship between Twitter activity (i.e., the number of unique tweets about an article) and number of citations. Twitter activity was a more important predictor of citation rates than 5-year journal impact factor. Moreover, Twitter activity was not driven by journal impact factor; the ‘highest-impact’ journals were not necessarily the most discussed online. The effect of Twitter activity was only about a fifth as strong as time since publication; accounting for this confounding factor was critical for estimating the true effects of Twitter use. Articles in impactful journals can become heavily cited, but articles in journals with lower impact factors can generate considerable Twitter activity and also become heavily cited. Authors may benefit from establishing a strong social media presence, but should not expect research to become highly cited solely through social media promotion. Our research demonstrates that altmetrics and traditional metrics can be closely related, but not identical. We suggest that both altmetrics and traditional citation rates can be useful metrics of research impact.

show abstract

“…A previous study using simulations had shown that negative binomial regression had a tendency to identify non-existent relationships at a rate above the significance level set, showing that conclusions drawn from negative binomial regression are unsafe (Thelwall & Wilson, 2014b). Whilst this conclusion was not confirmed by the analysis of Information Research articles and knowledge management articles (Ajiferuke & Famoye, 2015), the number of dependant variable tested was too small and the nature of the datasets tested too restricted to give convincing evidence and so the use of negative binomial regression for citation data remains problematic.…”

Section: Introductionmentioning

confidence: 94%

The discretised lognormal and hooked power law distributions for complete citation data: Best options for modelling and regression

Thelwall

2016

Journal of Informetrics

View full text Add to dashboard Cite

Identifying the statistical distribution that best fits citation data is important to allow robust and powerful quantitative analyses. Whilst previous studies have suggested that both the hooked power law and discretised lognormal distributions fit better than the power law and negative binomial distributions, no comparisons so far have covered all articles within a discipline, including those that are uncited. Based on an analysis of 26 different Scopus subject areas in seven different years, this article reports comparisons of the discretised lognormal and the hooked power law with citation data, adding 1 to citation counts in order to include zeros. The hooked power law fits better in two thirds of the subject/year combinations tested for journal articles that are at least three years old, including most medical, life and natural sciences, and for virtually all subject areas for younger articles. Conversely, the discretised lognormal tends to fit best for arts, humanities, social science and engineering fields. The difference between the fits of the distributions is mostly small, however, and so either could reasonably be used for modelling citation data. For regression analyses, however, the best option is to use ordinary least squares regression applied to the natural logarithm of citation counts plus one, especially for sets of younger articles, because of the increased precision of the parameters.

show abstract

Modelling count response variables in informetric studies: Comparison among count, linear, and lognormal regression models

Cited by 42 publications

References 46 publications

The effect of bioRxiv preprints on citations and altmetrics

The effect of bioRxiv preprints on citations and altmetrics

Twitter Predicts Citation Rates of Ecological Research

The discretised lognormal and hooked power law distributions for complete citation data: Best options for modelling and regression

Contact Info

Product

Resources

About