Abstract
First, the latent class (LC) model and its application to the Tendency
Survey are reviewed. Then we go on to the main focus in the paper: the application of the
latent Markov (LM) model to capture the dynamics in the Tendency Survey. The classes of the
static LC model are now seen as states in a dynamic model that allow individual (firms) to change
their state from one time to the next according to a Markov process. Stationary as well as nonstationary
models are admissible. Finally, in the concluding section we point to a more general class
of Markov models - the Mixed Latent Markov (MLM) with time-constant and time-varying covariates
- that offers a framework for Markov modelling that can be utilised in the further analyses of
the Tendency Survey.
Survey are reviewed. Then we go on to the main focus in the paper: the application of the
latent Markov (LM) model to capture the dynamics in the Tendency Survey. The classes of the
static LC model are now seen as states in a dynamic model that allow individual (firms) to change
their state from one time to the next according to a Markov process. Stationary as well as nonstationary
models are admissible. Finally, in the concluding section we point to a more general class
of Markov models - the Mixed Latent Markov (MLM) with time-constant and time-varying covariates
- that offers a framework for Markov modelling that can be utilised in the further analyses of
the Tendency Survey.
| Originalsprog | Engelsk |
|---|---|
| Titel | Symposium i anvendt statistik |
| Redaktører | Peter Linde |
| Forlag | Danmarks Statistik |
| Publikationsdato | 2015 |
| Udgave | 2015 |
| Sider | 99-113 |
| ISBN (Trykt) | 978-87-501-2171-8 |
| Status | Udgivet - 2015 |
| Begivenhed | Symposium i Anvendt Statistik - Danmarks Tekniske Universitet, Danmark Varighed: 26 jan. 2015 → 28 jan. 2015 Konferencens nummer: 37 |
Konference
| Konference | Symposium i Anvendt Statistik |
|---|---|
| Nummer | 37 |
| Lokation | Danmarks Tekniske Universitet |
| Land/Område | Danmark |
| Periode | 26/01/2015 → 28/01/2015 |