Theil proposed three instruments to be applied in regression analysis with prior stochastic information. Two of them are: a quadratic form (known as compatibility test) and a measure of the share of sample information on posterior variance. Duly modified, they are proposed in the paper to investigate if data come from a linear regression model or from a linear latent growth model. We compare a covariance matrix due to sample information and a total variance due to both sample and prior information following the two approaches: (i) by the difference of two quadratic forms based on the structure of the Theil’s compatibility test and (ii) by the product of two covariance matrices based on a measure of an estimated share of sample information in total variability. The first approach depends on the approximations to chi-square distributions which are discussed in the Appendix. The second one is based on a comparison of empirical cumulative distribution functions. A simple algorithm based on the bootstrap is proposed for this comparison.
SOME PROPOSALS OF THEIL USED TO DISCRIMINATE BETWEEN A LINEAR LATENT GROWTH MODEL AND A LINEAR REGRESSION MODEL / Marco Barnabani. - In: FAR EAST JOURNAL OF THEORETICAL STATISTICS. - ISSN 0972-0863. - STAMPA. - (2014), pp. 19-40.
SOME PROPOSALS OF THEIL USED TO DISCRIMINATE BETWEEN A LINEAR LATENT GROWTH MODEL AND A LINEAR REGRESSION MODEL
BARNABANI, MARCO
2014
Abstract
Theil proposed three instruments to be applied in regression analysis with prior stochastic information. Two of them are: a quadratic form (known as compatibility test) and a measure of the share of sample information on posterior variance. Duly modified, they are proposed in the paper to investigate if data come from a linear regression model or from a linear latent growth model. We compare a covariance matrix due to sample information and a total variance due to both sample and prior information following the two approaches: (i) by the difference of two quadratic forms based on the structure of the Theil’s compatibility test and (ii) by the product of two covariance matrices based on a measure of an estimated share of sample information in total variability. The first approach depends on the approximations to chi-square distributions which are discussed in the Appendix. The second one is based on a comparison of empirical cumulative distribution functions. A simple algorithm based on the bootstrap is proposed for this comparison.File | Dimensione | Formato | |
---|---|---|---|
PaperBarnabani.pdf
accesso aperto
Descrizione: Articolo principale
Tipologia:
Pdf editoriale (Version of record)
Licenza:
Open Access
Dimensione
1.45 MB
Formato
Adobe PDF
|
1.45 MB | Adobe PDF |
I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.