A Review of Spatial Probit Models: Estimation, Model Selection and Applications focused on Human Behavior

Estás aquí:
  1. Home
  2. A Review of Spatial Probit Models: Estimation, Model Selection and Applications focused on Human Behavior
  • Autor: MIguel Ángel de la Llave Montiel
  • Director/es: Fernando López Hernández
  • Defensa: 28/6/2022 - Universidad Nacional de Educación a Distancia
  • Tribunal: Mariano Matilla García, Manuel Ruiz Marín, Gema Fernández-Avilés Calderón
  • Calificación: Sobresaliente cum laude
  • Ver publicaciones relacionadas

Resumen:

Esta Tesis Doctoral la conforman cuatro estudios independientes enmarcados dentro de la temática del modelo probit espacial.
En el primer capítulo mostramos la evolución del conjunto de técnicas econométricas espaciales centrándonos en el probit espacial. Su estimación es un proceso complejo. La interdependencia de las observaciones conlleva la no esfericidad de los residuos y el problema conduce a maximizar el logaritmo de una distribución multivariante N-dimensional. Se desarrollan las metodologías propuestas hasta el momento:(Expectation-Maximization(EM), Generalización de los momentos(GMM), Gibbs Bayesian Sampling(Gibbs), Recursive Importance Sampling(RIS), GMM Lineal(LGMM) y Máxima Verosimilitud Aproximada(ML)).
Entre las aportaciones más relevantes se encuentra el cálculo de la precisión en las estimaciones de dichos algoritmos a través de un ejercicio de Monte Carlo ante modelos autorregresivos espaciales. La principal diferencia entre los algoritmos es en el sesgo de los coeficientes y en el coste computacional. Destacan Gibbs y ML, proporcionando ambos una buena precisión y solo distinguiéndose en que Gibbs tiene un proceso de convergencia ligeramente más lento.
Otra aportación es la evaluación de la relevancia del factor espacial dentro de la investigación en la modelización del comportamiento humano, concluyendo que aún no tiene un peso muy relevante.
El segundo capítulo resolvemos el problema de fuga de clientes a través de un modelo autorregresivo espacial. El estudio se centra en resolver principalmente la correlación entre un set de datos y la variable endógena dicotómica encontrando una forma funcional idónea. El coeficiente autorregresivo del modelo final nos dice que un porcentaje significativo de la probabilidad final del cliente de abandonar la compañía viene por efectos marginales indirectos. Dentro del modelo espacial, introducimos dentro del proceso de estimación una fase de deslinealización a través de la técnica de Multivariate Adaptive Regression Splines(MARS). El modelo espacial final resultante del análisis muestra mejores estadísticos que el modelo clásico tanto en precisión como en idoneidad de los residuos.
En el tercer capítulo profundizamos en los factores que mueven al usuario de una aplicación de compra por internet a permanecer inactivo por una larga duración. El modelo probit espacial mejora los resultados del modelo clásico y por lo tanto es una evidencia más sobre los beneficios de este tipo de modelización y su empleabilidad en la gestión de negocios. El comportamiento de la inactividad de clientes cercanos se demuestra estar codeterminado, detectándose conductas miméticas entre ellos. La especificación se ha mejorado con la búsqueda de patrones no-lineales con modelos generales aditivos(GAM) y con la incorporación de variables exógenas retardadas espacialmente.
En el cuarto capítulo analizamos las consecuencias de una incorrecta especificación en un modelo probit espacial y abordamos la búsqueda de la correcta selección del probit. La principal aportación es la configuración de dos algoritmos de selección de la verdadera especificación del modelo probit espacial. Proponemos una estrategia de lo específico a lo general(Stge) y otra de lo general a lo específico(Gets). La comparativa entre ambas técnicas la realizamos a través de una simulación de Monte Carlo para 5 tipos de especificaciones reales: Modelo Independiente(SIM), Modelo Autorregresivo Espacial(SAR), Modelo de Dependencia Espacial en el Error(SEM), Modelo con variable retardada espacialmente(SLX) y Modelo autorregresivo con variable retardada espacialmente-Durbin(SDM). Las estrategias de selección presentan un rendimiento superior al 85% de casos correctamente seleccionados. Resulta difícil decidir contundentemente qué estrategia es la mejor. Bajo condiciones ideales Stge funciona ligeramente mejor que Gets. Sin embargo, cuando introducimos situaciones no ideales como entonces hay ciertas ocasiones que Gets es más precisa.

Publicado el
WP_Query Object ( [query] => Array ( [post_type] => publicaciones [meta_key] => _wpcf_belongs_tesis_id [meta_value] => 3418 ) [query_vars] => Array ( [post_type] => publicaciones [meta_key] => _wpcf_belongs_tesis_id [meta_value] => 3418 [error] => [m] => [p] => 0 [post_parent] => [subpost] => [subpost_id] => [attachment] => [attachment_id] => 0 [name] => [pagename] => [page_id] => 0 [second] => [minute] => [hour] => [day] => 0 [monthnum] => 0 [year] => 0 [w] => 0 [category_name] => [tag] => [cat] => [tag_id] => [author] => [author_name] => [feed] => [tb] => [paged] => 0 [preview] => [s] => [sentence] => [title] => [fields] => [menu_order] => [embed] => [category__in] => Array ( ) [category__not_in] => Array ( ) [category__and] => Array ( ) [post__in] => Array ( ) [post__not_in] => Array ( ) [post_name__in] => Array ( ) [tag__in] => Array ( ) [tag__not_in] => Array ( ) [tag__and] => Array ( ) [tag_slug__in] => Array ( ) [tag_slug__and] => Array ( ) [post_parent__in] => Array ( ) [post_parent__not_in] => Array ( ) [author__in] => Array ( ) [author__not_in] => Array ( ) [ignore_sticky_posts] => [suppress_filters] => [cache_results] => 1 [update_post_term_cache] => 1 [lazy_load_term_meta] => 1 [update_post_meta_cache] => 1 [posts_per_page] => 10 [nopaging] => [comments_per_page] => 50 [no_found_rows] => [order] => DESC ) [tax_query] => WP_Tax_Query Object ( [queries] => Array ( ) [relation] => AND [table_aliases:protected] => Array ( ) [queried_terms] => Array ( ) [primary_table] => wp_posts [primary_id_column] => ID ) [meta_query] => WP_Meta_Query Object ( [queries] => Array ( [0] => Array ( [key] => _wpcf_belongs_tesis_id [value] => 3418 ) [relation] => OR ) [relation] => AND [meta_table] => wp_postmeta [meta_id_column] => post_id [primary_table] => wp_posts [primary_id_column] => ID [table_aliases:protected] => Array ( [0] => wp_postmeta ) [clauses:protected] => Array ( [wp_postmeta] => Array ( [key] => _wpcf_belongs_tesis_id [value] => 3418 [compare] => = [compare_key] => = [alias] => wp_postmeta [cast] => CHAR ) ) [has_or_relation:protected] => ) [date_query] => [request] => SELECT SQL_CALC_FOUND_ROWS wp_posts.ID FROM wp_posts INNER JOIN wp_postmeta ON ( wp_posts.ID = wp_postmeta.post_id ) WHERE 1=1 AND ( ( wp_postmeta.meta_key = '_wpcf_belongs_tesis_id' AND wp_postmeta.meta_value = '3418' ) ) AND ((wp_posts.post_type = 'publicaciones' AND (wp_posts.post_status = 'publish'))) GROUP BY wp_posts.ID ORDER BY wp_posts.post_date DESC LIMIT 0, 10 [posts] => Array ( [0] => WP_Post Object ( [ID] => 3384 [post_author] => 19 [post_date] => 2022-05-24 12:37:27 [post_date_gmt] => 2022-05-24 10:37:27 [post_content] => Abstract Geography has previously been noted as a decisive factor in business literature. This paper provides evidence of the significant role geography plays in customer lapse behaviour in an urban environment. This novel approach is based on the idea that the customers who cancel all policies and leave the company are not randomly distributed; rather, a mimetic performance of close individuals is noted. The physical proximity of the customer to the geographical focus (strategical centre, as insurance offices) and the interaction with nearby customer are spatial factors that increase (or decrease) the probability of churning. An empirical analysis using more than 7000 spatially georeferenced offline customers of a Spanish insurance company in the urban area of Madrid (Spain) demonstrated that the customer’s proximity to offices of such insurance company under study decreases the probability of churning, whereas high lapse risk was detected in customers in the surroundings of the company’s competitor branches. In addition, we identified spatial autocorrelation in churn probability, thus demonstrating that the probability of churn of a customer increases if nearby customers churn. [post_title] => The impact of geographical factors on churn prediction: an application to an insurance company in Madrid's urban area [post_excerpt] => [post_status] => publish [comment_status] => open [ping_status] => open [post_password] => [post_name] => the-impact-of-geographical-factors-on-churn-prediction-an-application-to-an-insurance-company-in-madrids-urban-area [to_ping] => [pinged] => [post_modified] => 2022-07-01 16:26:39 [post_modified_gmt] => 2022-07-01 14:26:39 [post_content_filtered] => [post_parent] => 0 [guid] => http://doctoradodecide.com/?post_type=publicaciones&p=3384 [menu_order] => 0 [post_type] => publicaciones [post_mime_type] => [comment_count] => 0 [filter] => raw ) [1] => WP_Post Object ( [ID] => 3382 [post_author] => 19 [post_date] => 2022-05-24 12:33:51 [post_date_gmt] => 2022-05-24 10:33:51 [post_content] => Abstract This paper presents evidence of the significant role that geography plays in customer churn behaviour in online retail. In an urban environment, mimetic behaviours are found to affect nearby individuals. This novel approach is based on the idea that customer churn is not randomly distributed across the map. This paper analyses more than 2,000 spatially georeferenced customers and demonstrates that customers show different patterns when deciding to cease activity, and that other factors besides spatial autocorrelation influence churn probability. Finally, the results prove that including spatial spillover in models improves predictability. This improvement results in substantial economic benefits since marketing managers can consequently reduce their company's loss of customers more effectively [post_title] => Spatial models for online retail churn: Evidence from an online grocery delivery service in Madrid [post_excerpt] => [post_status] => publish [comment_status] => open [ping_status] => open [post_password] => [post_name] => spatial-models-for-online-retail-churn-evidence-from-an-online-grocery-delivery-service-in-madrid [to_ping] => [pinged] => [post_modified] => 2022-07-01 16:25:38 [post_modified_gmt] => 2022-07-01 14:25:38 [post_content_filtered] => [post_parent] => 0 [guid] => http://doctoradodecide.com/?post_type=publicaciones&p=3382 [menu_order] => 0 [post_type] => publicaciones [post_mime_type] => [comment_count] => 0 [filter] => raw ) ) [post_count] => 2 [current_post] => -1 [in_the_loop] => [post] => WP_Post Object ( [ID] => 3384 [post_author] => 19 [post_date] => 2022-05-24 12:37:27 [post_date_gmt] => 2022-05-24 10:37:27 [post_content] => Abstract Geography has previously been noted as a decisive factor in business literature. This paper provides evidence of the significant role geography plays in customer lapse behaviour in an urban environment. This novel approach is based on the idea that the customers who cancel all policies and leave the company are not randomly distributed; rather, a mimetic performance of close individuals is noted. The physical proximity of the customer to the geographical focus (strategical centre, as insurance offices) and the interaction with nearby customer are spatial factors that increase (or decrease) the probability of churning. An empirical analysis using more than 7000 spatially georeferenced offline customers of a Spanish insurance company in the urban area of Madrid (Spain) demonstrated that the customer’s proximity to offices of such insurance company under study decreases the probability of churning, whereas high lapse risk was detected in customers in the surroundings of the company’s competitor branches. In addition, we identified spatial autocorrelation in churn probability, thus demonstrating that the probability of churn of a customer increases if nearby customers churn. [post_title] => The impact of geographical factors on churn prediction: an application to an insurance company in Madrid's urban area [post_excerpt] => [post_status] => publish [comment_status] => open [ping_status] => open [post_password] => [post_name] => the-impact-of-geographical-factors-on-churn-prediction-an-application-to-an-insurance-company-in-madrids-urban-area [to_ping] => [pinged] => [post_modified] => 2022-07-01 16:26:39 [post_modified_gmt] => 2022-07-01 14:26:39 [post_content_filtered] => [post_parent] => 0 [guid] => http://doctoradodecide.com/?post_type=publicaciones&p=3384 [menu_order] => 0 [post_type] => publicaciones [post_mime_type] => [comment_count] => 0 [filter] => raw ) [comment_count] => 0 [current_comment] => -1 [found_posts] => 2 [max_num_pages] => 1 [max_num_comment_pages] => 0 [is_single] => [is_preview] => [is_page] => [is_archive] => 1 [is_date] => [is_year] => [is_month] => [is_day] => [is_time] => [is_author] => [is_category] => [is_tag] => [is_tax] => [is_search] => [is_feed] => [is_comment_feed] => [is_trackback] => [is_home] => [is_privacy_policy] => [is_404] => [is_embed] => [is_paged] => [is_admin] => [is_attachment] => [is_singular] => [is_robots] => [is_favicon] => [is_posts_page] => [is_post_type_archive] => 1 [query_vars_hash:WP_Query:private] => addc27109fb0be4ed650d39befaae13c [query_vars_changed:WP_Query:private] => [thumbnails_cached] => [allow_query_attachment_by_filename:protected] => [stopwords:WP_Query:private] => [compat_fields:WP_Query:private] => Array ( [0] => query_vars_hash [1] => query_vars_changed ) [compat_methods:WP_Query:private] => Array ( [0] => init_query_flags [1] => parse_tax_query ) )

Publicaciones relacionadas

Array ( [_edit_lock] => Array ( [0] => 1665211055:19 ) [_edit_last] => Array ( [0] => 19 ) [_wpcf-autor-sort-order] => Array ( [0] => a:1:{i:0;i:24951;} ) [wpcf-autor] => Array ( [0] => MIguel Ángel de la Llave Montiel ) [wpcf-director] => Array ( [0] => Fernando López Hernández ) [wpcf-fecha-defensa] => Array ( [0] => 1656374400 ) [wpcf-lugar] => Array ( [0] => Universidad Nacional de Educación a Distancia ) [wpcf-tribunal] => Array ( [0] => Mariano Matilla García Manuel Ruiz Marín Gema Fernández-Avilés Calderón ) [wpcf-calificacion] => Array ( [0] => 10 ) )