-
Evaluation of Logistic Regression Applied to Respondent-Driven Samples: Simulated and Real Data
Authors:
Sandro Sperandei,
Leonardo S. Bastos,
Marcelo Ribeiro-Alves,
Arianne Reis,
Francisco I. Bastos
Abstract:
Objective: To investigate the impact of different logistic regression estimators applied to RDS samples obtained by simulation and real data. Methods: Four simulated populations were created combining different connectivity models, levels of clusterization and infection processes. Each subject in the population received two attributes, only one of them related to the infection process. From each p…
▽ More
Objective: To investigate the impact of different logistic regression estimators applied to RDS samples obtained by simulation and real data. Methods: Four simulated populations were created combining different connectivity models, levels of clusterization and infection processes. Each subject in the population received two attributes, only one of them related to the infection process. From each population, RDS samples with different sizes were obtained. Similarly, RDS samples were obtained from a real-world dataset. Three logistic regression estimators were applied to assess the association between the attributes and the infection status, and subsequently the observed coverage of each was measured. Results: The type of connectivity had more impact on estimators performance than the clusterization level. In simulated datasets, unweighted logistic regression estimators emerged as the best option, although all estimators showed a fairly good performance. In the real dataset, the performance of weighted estimators presented some instabilities, making them a risky option. Conclusion: An unweighted logistic regression estimator is a reliable option to be applied to RDS samples, with similar performance to random samples and, therefore, should be the preferred option.
△ Less
Submitted 11 January, 2021;
originally announced January 2021.
-
Fast approaches for Bayesian estimation of size of hard-to-reach populations using Network Scale-up
Authors:
Leonardo S Bastos,
Natalia S Paiva,
Francisco I Bastos,
Daniel A M Villela
Abstract:
The Network scale-up method is commonly used to overcome difficulties in estimating the size of hard-to-reach populations. The method uses indirect information based on social network of each participant taken from the general population, but in some applications a fast computational approach would be highly recommended. We propose a Gibbs sampling method and a Monte Carlo approach to sample from…
▽ More
The Network scale-up method is commonly used to overcome difficulties in estimating the size of hard-to-reach populations. The method uses indirect information based on social network of each participant taken from the general population, but in some applications a fast computational approach would be highly recommended. We propose a Gibbs sampling method and a Monte Carlo approach to sample from the random degree model. We applied the abovementioned analytical strategies to previous data on heavy drug users from Curitiba, Brazil.
△ Less
Submitted 12 April, 2018;
originally announced April 2018.
-
Binary regression analysis with network structure of respondent-driven sampling data
Authors:
Leonardo S. Bastos,
Adriana A. Pinho,
Claudia Codeço,
Francisco I. Bastos
Abstract:
Respondent-driven sampling (RDS) is a procedure to sample from hard-to-reach populations. It has been widely used in several countries, especially in the monitoring of HIV/AIDS and other sexually transmitted infections. Hard-to-reach populations have had a key role in the dynamics of such epidemics and must inform evidence-based initiatives aiming to curb their spread. In this paper, we present a…
▽ More
Respondent-driven sampling (RDS) is a procedure to sample from hard-to-reach populations. It has been widely used in several countries, especially in the monitoring of HIV/AIDS and other sexually transmitted infections. Hard-to-reach populations have had a key role in the dynamics of such epidemics and must inform evidence-based initiatives aiming to curb their spread. In this paper, we present a simple test for network dependence for a binary response variable. We estimate the prevalence of the response variable. We also propose a binary regression model taking into account the RDS structure which is included in the model through a latent random effect with a correlation structure. The proposed model is illustrated in a RDS study for HIV and Syphilis in men who have sex with men implemented in Campinas (Brazil).
△ Less
Submitted 25 June, 2012;
originally announced June 2012.
