Hidden Markov Models (HMMs) are highly flexible probabilistic models introduced in the 1960s. They became popular in time series analysis due to their ability to effectively model complex temporal dynamics. In recent years, their use has also expanded to longitudinal data analysis. Although the application of HMMs to such data was first proposed in the 1970s, interest in and utilization of these models have grown significantly only recently, thanks to advancements in computational power and methodological developments.In this context, HMMs stand out for their ability to identify, based on a set of observed variables, distinct profiles of unobservable behavior, enabling the identification of groups of individuals with similar behavioral patterns over time. This type of analysis is particularly relevant in fields such as social, behavioral, and medical research, as individuals often exhibit a set of behaviors influenced by latent factors. The objective of this thesis is to provide new evidence on the applicability and potential of Hidden Markov Models in the analysis of longitudinal data within an applied context such as mental health. Indeed, their use in psychology, and particularly in the study of depression, remains relatively limited compared to other fields, such as biomedicine or economics.
I modelli di Markov nascosti (Hidden Markov Models, HMMs) sono modelli probabilistici molto flessibili, introdotti negli anni ’60 e divenuti popolari nell’analisi delle serie storiche, grazie alla loro capacità di modellare efficacemente dinamiche temporali complesse. Negli ultimi anni il loro utilizzo si è diffuso anche nell’analisi di dati longitudinali; sebbene l’applicazione degli HMMs a tali dati sia stata proposta per la prima volta negli anni ’70, l’interesse e il loro utilizzo sono aumentati notevolmente solo di recente, grazie al potenziamento delle capacità computazionali e ai recenti sviluppi metodologici. In questo contesto, gli HMMs si distinguono per la capacità di caratterizzare, a partire da un insieme di variabili osservate, diversi profili di comportamento non direttamente osservabili, rendendo possibile l’identificazione di gruppi di individui con pattern di comportamento simili nel tempo. Tale analisi è di particolare interesse ad esempio nella ricerca sociale, comportamentale e medica, perchè i soggetti possono manifestare un insieme di comportamenti dovuti, spesso, a fattori latenti. Obiettivo di questa tesi è dunque fornire nuova evidenza dell’applicabilità e delle potenzialità dei modelli di Markov nascosti nell’analisi di dati longitudinali, in un contesto applicativo come quello della salute mentale, attraverso un’analisi sulla depressione negli anziani in Italia. Infatti, il loro utilizzo in ambito psicologico, e in particolare nello studio della depressione, è ancora relativamente limitato rispetto ad altri campi come quello biomedico o economico.
Uno studio longitudinale sulla depressione negli anziani tramite modelli di Markov nascosti
PILEGGI, SILVIA
2024/2025
Abstract
Hidden Markov Models (HMMs) are highly flexible probabilistic models introduced in the 1960s. They became popular in time series analysis due to their ability to effectively model complex temporal dynamics. In recent years, their use has also expanded to longitudinal data analysis. Although the application of HMMs to such data was first proposed in the 1970s, interest in and utilization of these models have grown significantly only recently, thanks to advancements in computational power and methodological developments.In this context, HMMs stand out for their ability to identify, based on a set of observed variables, distinct profiles of unobservable behavior, enabling the identification of groups of individuals with similar behavioral patterns over time. This type of analysis is particularly relevant in fields such as social, behavioral, and medical research, as individuals often exhibit a set of behaviors influenced by latent factors. The objective of this thesis is to provide new evidence on the applicability and potential of Hidden Markov Models in the analysis of longitudinal data within an applied context such as mental health. Indeed, their use in psychology, and particularly in the study of depression, remains relatively limited compared to other fields, such as biomedicine or economics.File | Dimensione | Formato | |
---|---|---|---|
PILEGGI_SILVIA.pdf
accesso aperto
Dimensione
864.3 kB
Formato
Adobe PDF
|
864.3 kB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/84091