Samson Yaekob Assele , Michel Meulders , Martina Vandebroek
{"title":"利用设计特征进行离散选择实验的样本量选择","authors":"Samson Yaekob Assele , Michel Meulders , Martina Vandebroek","doi":"10.1016/j.jocm.2023.100436","DOIUrl":null,"url":null,"abstract":"<div><p>In discrete choice experiment (DCE) studies, selecting the appropriate sample size remains a challenge. The question of the required sample size for a DCE is addressed in the literature in two distinct approaches: a rule-of-thumb approach and an approach based on the statistical error of the parameter of interest. The former is less accurate and does not depend on the desired power and significance level, whereas the latter requires knowing the complete design which may not be known at the planning stage. This paper proposes a new rule of thumb as well as a new regression-based method that requires knowing certain design characteristics rather than the complete design and takes into account the power and significance level. We compare the sample size estimated using the proposed methods with the true required sample size based on the statistical error of the parameter of interest and the approximations given by the existing rules of thumb. The results show that both the new rule of thumb and the regression-based approach improve the magnitude and proportion of underestimation compared to the most commonly used rule of thumb of Orme. Though the proposed approaches perform in general similarly to Tang’s rule which improves Orme’s rule, they seem to do better for large settings in terms of the number of choice sets and the number of alternatives per choice set in reducing underestimation. Moreover, we have demonstrated the possibility to adapt the regression-based approaches to take into account other scenarios and choice set complexity.</p></div>","PeriodicalId":46863,"journal":{"name":"Journal of Choice Modelling","volume":"49 ","pages":"Article 100436"},"PeriodicalIF":2.8000,"publicationDate":"2023-09-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Sample size selection for discrete choice experiments using design features\",\"authors\":\"Samson Yaekob Assele , Michel Meulders , Martina Vandebroek\",\"doi\":\"10.1016/j.jocm.2023.100436\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><p>In discrete choice experiment (DCE) studies, selecting the appropriate sample size remains a challenge. The question of the required sample size for a DCE is addressed in the literature in two distinct approaches: a rule-of-thumb approach and an approach based on the statistical error of the parameter of interest. The former is less accurate and does not depend on the desired power and significance level, whereas the latter requires knowing the complete design which may not be known at the planning stage. This paper proposes a new rule of thumb as well as a new regression-based method that requires knowing certain design characteristics rather than the complete design and takes into account the power and significance level. We compare the sample size estimated using the proposed methods with the true required sample size based on the statistical error of the parameter of interest and the approximations given by the existing rules of thumb. The results show that both the new rule of thumb and the regression-based approach improve the magnitude and proportion of underestimation compared to the most commonly used rule of thumb of Orme. Though the proposed approaches perform in general similarly to Tang’s rule which improves Orme’s rule, they seem to do better for large settings in terms of the number of choice sets and the number of alternatives per choice set in reducing underestimation. Moreover, we have demonstrated the possibility to adapt the regression-based approaches to take into account other scenarios and choice set complexity.</p></div>\",\"PeriodicalId\":46863,\"journal\":{\"name\":\"Journal of Choice Modelling\",\"volume\":\"49 \",\"pages\":\"Article 100436\"},\"PeriodicalIF\":2.8000,\"publicationDate\":\"2023-09-08\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Choice Modelling\",\"FirstCategoryId\":\"96\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S1755534523000374\",\"RegionNum\":3,\"RegionCategory\":\"经济学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"ECONOMICS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Choice Modelling","FirstCategoryId":"96","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1755534523000374","RegionNum":3,"RegionCategory":"经济学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ECONOMICS","Score":null,"Total":0}
Sample size selection for discrete choice experiments using design features
In discrete choice experiment (DCE) studies, selecting the appropriate sample size remains a challenge. The question of the required sample size for a DCE is addressed in the literature in two distinct approaches: a rule-of-thumb approach and an approach based on the statistical error of the parameter of interest. The former is less accurate and does not depend on the desired power and significance level, whereas the latter requires knowing the complete design which may not be known at the planning stage. This paper proposes a new rule of thumb as well as a new regression-based method that requires knowing certain design characteristics rather than the complete design and takes into account the power and significance level. We compare the sample size estimated using the proposed methods with the true required sample size based on the statistical error of the parameter of interest and the approximations given by the existing rules of thumb. The results show that both the new rule of thumb and the regression-based approach improve the magnitude and proportion of underestimation compared to the most commonly used rule of thumb of Orme. Though the proposed approaches perform in general similarly to Tang’s rule which improves Orme’s rule, they seem to do better for large settings in terms of the number of choice sets and the number of alternatives per choice set in reducing underestimation. Moreover, we have demonstrated the possibility to adapt the regression-based approaches to take into account other scenarios and choice set complexity.