Estimating logit models with small samples

Rainey, Carlisle; McCaskey, Kelly

doi:10.1017/psrm.2021.9

Cited by 46 publications

(38 citation statements)

References 38 publications

(61 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Frequency, percentages, and cross-tabulation were used to describe the sample. The prevalence of cervical cancer screening was approximately 4% in the female population, rendering the dataset imbalanced or sparse [27][28][29]. Reliance on the Maximum Likelihood Estimation (MLE) based logistic regression will produce coefficients that are biased and unreliable [27][28][29].…”

Section: Discussionmentioning

confidence: 99%

“…The prevalence of cervical cancer screening was approximately 4% in the female population, rendering the dataset imbalanced or sparse [27][28][29]. Reliance on the Maximum Likelihood Estimation (MLE) based logistic regression will produce coefficients that are biased and unreliable [27][28][29]. In such situations, a data analyst has a choice among the exact logistic regression, penalized maximum likelihood estimation (PMLE), and the Firth logistic regression.…”

Section: Discussionmentioning

confidence: 99%

“…In such situations, a data analyst has a choice among the exact logistic regression, penalized maximum likelihood estimation (PMLE), and the Firth logistic regression. The exact logistic regression, although ideal, is computationally intensive [27,28]. The Firth logistic regression, although fairly new, can easily be computed in most statistical software.…”

Section: Discussionmentioning

confidence: 99%

“…In 1993, David Firth introduced a statistical method that corrects the MLE bias that is named after the author as Firth logistic regression [30]. Packages exist in multiple statistical software to perform this analysis, and it has been established that the Firth logistic regression produces accurate results more consistently [27][28][29]. For our study, we implemented the Firth logistic regression using the "firthlogit" command in STATA to perform the bivariate analyses between the outcome variable and each of the explanatory variables.…”

Section: Discussionmentioning

confidence: 99%

See 3 more Smart Citations

Cervical cancer screening prevalence and its correlates in Cameroon: secondary data analysis of the 2018 demographic and health surveys

et al. 2021

View full text Add to dashboard Cite

Background Cervical cancer, although preventable, is the fourth most common cancer among women globally, and the second most common and deadliest gynaecological cancer in low-and-middle-income countries. Screening is key to the prevention and early detection of the disease for treatment. A few studies estimated the prevalence of cervical cancer screening and its correlates in Cameroon but relied on data that were limited to certain regions of the country. Therefore, this study sought to examine the prevalence and correlates of cervical cancer screening among Cameroonian women using current data that is nationally representative of reproductive-age women. Methods We used secondary data from the 2018 Cameroon Demographic and Health Survey. Summary statistics were used for the sample description. We employed the Firth logistic regression using the “firthlogit” command in STATA-14 to perform the bivariate analyses between the outcome variable and each of the explanatory variables. Given that all the explanatory variables were statistically significant correlates, they were all adjusted for in a multivariable analysis. All analyses were performed in STATA version 14. Results The proportion of Cameroonian women who have ever screened for cervical cancer continue to remain low at approximately 4%. In the adjusted model, women with the following sociodemographic characteristics have a higher likelihood of undergoing cervical cancer screening: ever undergone HIV screening (AOR = 4.446, 95% CI: 2.475, 7.986), being 24–34 years (AOR = 2.233, 95% CI: 1.606, 3.103) or 35–44 years (AOR = 4.008, 95% CI: 2.840, 5.657) or at least 45 years old (AOR = 5.895, 95% CI: 3.957, 8.784), having attained a post-secondary education (AOR = 1.849, 95% CI: 1.032, 3.315), currently (AOR = 1.551, 95% CI: 1.177, 2.043) or previously married (AOR = 1.572, 95% CI: 1.073, 2.302), dwelling in the richest household (AOR = 4.139, 95% CI: 1.769, 9.682), and residing in an urban area (AOR = 1.403, 95% CI: 1.004,1.960). Except for the North-West region, residing in some five regions, compared to Yaounde, was negatively associated with cervical cancer screening. Conclusion Cervical cancer screening programs and policies should target Cameroonian women who are younger, less educated, and those in poor households and rural areas.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 2 more Smart Citations

Cervical cancer screening prevalence and its correlates in Cameroon: secondary data analysis of the 2018 demographic and health surveys

et al. 2021

View full text Add to dashboard Cite

show abstract

“…For logit models this is a relatively small sample size. Such relatively small sample sizes in the logit models could be problematic for producing unbiased estimates (Rainey and McCaskey, 2021). To account for this, models 3 and 4 use a Firth logistic regression model that reduces small sample bias (Rainey and McCaskey, 2021).…”

Section: Empirical Analysismentioning

confidence: 99%

Territorial wars and absolute outcomes

Crisher

2021

Research & Politics

View full text Add to dashboard Cite

Why do some wars end with an absolute outcome, with state death or regime change? I argue that we are more likely to see absolute outcomes when we have territorial disputes with the potential for credible commitment problems and asymmetric disputants. In the absence of credible commitment problems, disputes are less likely to recur, and states are unlikely to seek to absorb the opponent state or remove its government. Among more symmetric disputants, states cannot impose an absolute outcome, and we are more likely to see recurrent disputes in the face of credible commitment problems. Only in very asymmetric dyads are we likely to have both the required willingness and opportunity to impose absolute outcomes to attempt to solve a credible commitment problem over territorial conflict.

show abstract

An evidence‐based approach to identifying resting sites of Eurasian otterLutra lutrafrom camera‐trap and field‐sign data

Findlay

Briers

Ingledew³

et al. 2022

Wildlife Biology

View full text Add to dashboard Cite

Mammals' resting sites (dens) are important features of their ecology. Eurasian otter Lutra lutra resting sites are strictly protected by UK and European legislation and are ostensibly identified from associated field-signs. This legislation is difficult to apply given the poor understanding of resting sites coupled with the lack of evidence supporting a field-sign signature. We aimed to use camera-trap data to identify resting sites, investigate whether field-signs differed between resting and non-resting sites and describe behaviours recorded on camera-traps that are associated with resting. An evidence-based approach to identifying resting sites of Eurasian otter Lutra lutra from camera-trap and field-sign data camera-trap data showed that otters frequently visited potential resting sites, characterised by a very short time within the structure (often < 4 min). Resting sites were characterised by longer durations (often hours) during the daytime and night-time. Based on these data, six of our 26 sites were identified as resting sites. Modelling suggested that no single field-sign had a clear association with resting sites. However, we found a hitherto unrecognised distinction between otter latrines (defecation sites) and spraint (scent-marking) sites, and that camera-trap observations of latrine behaviour and bedding collection were exclusive to resting sites. As bedding and latrines are not always visible, presence of either indicates a resting site but no interpretation can be drawn from their absence, so camera-trapping would be recommended to identify resting site status. Data simulations found that cameratrapping for 38 d in winter, followed by 38 d in spring, was the optimal approach for a 95% chance of detecting a rest across all resting sites. Ours is the first study to identify standards and expectations for surveys using camera-trap and field-signs at Eurasian otter resting sites. Our novel account of their resting activity facilitates better interpretation of legislation.

show abstract

Estimating logit models with small samples

Cited by 46 publications

References 38 publications

Cervical cancer screening prevalence and its correlates in Cameroon: secondary data analysis of the 2018 demographic and health surveys

Cervical cancer screening prevalence and its correlates in Cameroon: secondary data analysis of the 2018 demographic and health surveys

Territorial wars and absolute outcomes

An evidence‐based approach to identifying resting sites of Eurasian otterLutra lutrafrom camera‐trap and field‐sign data

Contact Info

Product

Resources

About