دانلود رایگان مقاله لاتین مقایسه گزینه ها برای عدم همبستگی در مدل بازاریابی مستقیم از سایت الزویر
عنوان فارسی مقاله:
مقایسه جایگزین ها برای عدم همبستگی نامتعارف در مدل های بازاریابی مستقیم
عنوان انگلیسی مقاله:
Comparing alternatives to account for unobserved heterogeneity in direct marketing models
سال انتشار : 2017
بخشی از مقاله انگلیسی:
2. Findings on unobserved heterogeneity in direct marketing models
In the following, we report results from past research on effects of direct mailings and unobserved heterogeneity. The first group of authors incorporates unobserved heterogeneity by finite mixture models. This group of authors tests whether a finite mixture model is superior to a homogenous model. Gon¨ul et ¨ al. (2000) specify a Box-Cox hazard function with various predictors. Whereas they assume predictors like gender and average consumption rate to be fixed, the baseline hazard together with predictors capturing the elapsed time since the last mailing and the accumulated number of mailings since the last purchase are assumed to be heterogeneous. They estimate several models starting with a one-segment solution. Based on the Bayesian Information Criterion (BIC), they find that a two-segment solution (consisting of 62% vs. 38% of the customers) performs best. Whereas they observe a wearout effect for the customers belonging to the larger segment, recent mailings increase the response rate for customers from both segments. Hruschka, Baumgartner, and Semmler (2003) also investigate the effect of mailings (in terms of number of mailings) and other variables on response behavior, i.e., response probability and sales. When they validate their semi-log and logistic response models, it turns out that one-segment, i.e., homogeneous, solutions are superior to two-segment solutions. They find positive relationships between number of mailings and response probability and sales, respectively. The following two studies refer to authors who employ MDP, which automatically results in the optimal number of segments. Hruschka (2010) estimates a semi-log response model with number of catalogs as one of the predictors. In addition, he compares a policy function to an instrumental variable model. Results based on cross validation predictive densities show that an instrumental variable model with on average 14 segments of customers performs better than the policy function model with on average 31 segments of customers. The predictor number of catalogs has a positive impact on sales. The study of Ansari and Mela (2003) examines the relationship of different characteristics of online mailings, i.e., e-mails, and the probability of clicking on one of the links in the e-mails. Predictors consist of e-mail characteristics, such as, e.g., the number and order of links in an e-mail. In addition, also the editorial content is being captured. They compare a MDP with a continuous mixture approach. Based on Pseudo Bayes factors (PsBF) they find that the MDP with on average 61 segments outperforms the continuous mixture. Findings regarding their predictors include a decreasing effectiveness of links when they appear later in the e-mail. The number of links on the other hand does not affect the clicking probability.
Comparing Explanations of Fertility Decline Using Event History ... https://www.stat.washington.edu/research/reports/1995/tr298.pdf by SM Lewis - 1995 - Cited by 5 - Related articles Nov 20, 1995 - Event History Models with Unobserved Heterogeneity ... Compound Laplace-Metropolis estimates were used to compute Bayes factors for comparing alternative models. The new methods enabled us to conclude that Iran's fertility ... 4.5 Accounting for Unobserved Heterogeneity Using MCMC : : : : : : : : : : 12. [PDF]What Is Heterogeneity? Why Is Heterogeneity Important? How Can ... https://www.ispor.org/congresses/Spain1111/presentations/w28_stulldonald.pdf Unobserved Heterogeneity: Innovative Approach. • Methods based on structural equation modeling (SEM) deal more efficiently with measuring unobserved heterogeneity. • Model-based methods have the advantage of using more rigorous approaches to compare alternative models. • Model-based methods can uncover ...