Details
| Translated title of the contribution | Statistical surveillance of spatial autoregressive processes with exogenous regressors |
|---|---|
| Original language | German |
| Pages (from-to) | 107-133 |
| Number of pages | 27 |
| Journal | AStA Wirtschafts- und Sozialstatistisches Archiv |
| Volume | 12 |
| Issue number | 2 |
| Publication status | Published - 1 Sept 2018 |
| Externally published | Yes |
Abstract
This paper deals with statistical process control of spatial autoregressive models with exogenous regressors. The main purpose is the extension of conventional methods of process control in time series analysis. These approaches are modified for applications of spatial monitoring. The method is illustrated by an example of social statistics dealing with natural as well as spatial population change regarding administrative districts of Germany. Via factor analysis latent variables are identified based on manifest variables, because independent factors are needed for the following analysis. Afterwards, the considered regions are divided into groups via cluster analysis. The results of cluster analysis helps to find a specific region of one cluster that is used for in-control estimation. The previously mentioned model is fitted to factor scores using the generalized method of moments. Multivariate control charts based on either exponential smoothing or cumulative sum are used to evaluate full-sample data regarding their control situation. Accordingly, we propose different approaches to sort the regions to be monitored. Eventually, the modified charts signalize structural changes regarding the model based on in-control data without permanent re-estimation.
ASJC Scopus subject areas
- Mathematics(all)
- Statistics and Probability
- Social Sciences(all)
- General Social Sciences
- Economics, Econometrics and Finance(all)
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: AStA Wirtschafts- und Sozialstatistisches Archiv, Vol. 12, No. 2, 01.09.2018, p. 107-133.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Verfahren zur Überwachung räumlicher autoregressiver Prozesse mit externen Regressoren
AU - Garthoff, Robert
AU - Otto, Philipp
N1 - Publisher Copyright: © 2018, Springer-Verlag GmbH Germany, part of Springer Nature.
PY - 2018/9/1
Y1 - 2018/9/1
N2 - This paper deals with statistical process control of spatial autoregressive models with exogenous regressors. The main purpose is the extension of conventional methods of process control in time series analysis. These approaches are modified for applications of spatial monitoring. The method is illustrated by an example of social statistics dealing with natural as well as spatial population change regarding administrative districts of Germany. Via factor analysis latent variables are identified based on manifest variables, because independent factors are needed for the following analysis. Afterwards, the considered regions are divided into groups via cluster analysis. The results of cluster analysis helps to find a specific region of one cluster that is used for in-control estimation. The previously mentioned model is fitted to factor scores using the generalized method of moments. Multivariate control charts based on either exponential smoothing or cumulative sum are used to evaluate full-sample data regarding their control situation. Accordingly, we propose different approaches to sort the regions to be monitored. Eventually, the modified charts signalize structural changes regarding the model based on in-control data without permanent re-estimation.
AB - This paper deals with statistical process control of spatial autoregressive models with exogenous regressors. The main purpose is the extension of conventional methods of process control in time series analysis. These approaches are modified for applications of spatial monitoring. The method is illustrated by an example of social statistics dealing with natural as well as spatial population change regarding administrative districts of Germany. Via factor analysis latent variables are identified based on manifest variables, because independent factors are needed for the following analysis. Afterwards, the considered regions are divided into groups via cluster analysis. The results of cluster analysis helps to find a specific region of one cluster that is used for in-control estimation. The previously mentioned model is fitted to factor scores using the generalized method of moments. Multivariate control charts based on either exponential smoothing or cumulative sum are used to evaluate full-sample data regarding their control situation. Accordingly, we propose different approaches to sort the regions to be monitored. Eventually, the modified charts signalize structural changes regarding the model based on in-control data without permanent re-estimation.
KW - Cluster analysis
KW - Demographic development
KW - Factor analysis
KW - Spatial autoregressive models
KW - Spatial process control
UR - http://www.scopus.com/inward/record.url?scp=85047266479&partnerID=8YFLogxK
U2 - 10.1007/s11943-018-0224-1
DO - 10.1007/s11943-018-0224-1
M3 - Artikel
AN - SCOPUS:85047266479
VL - 12
SP - 107
EP - 133
JO - AStA Wirtschafts- und Sozialstatistisches Archiv
JF - AStA Wirtschafts- und Sozialstatistisches Archiv
SN - 1863-8155
IS - 2
ER -