Options
Murphy, Anthony
Preferred name
Murphy, Anthony
Official Name
Murphy, Anthony
Research Output
Now showing 1 - 10 of 19
- PublicationA simple artificial regression based Lagrange multiplier test of normality in the probit modelA convenient artifical regression based LM test of non-normality in the probit model is derived using a Gram Charlier type A alternative. The test is simply derived and may be extended to the bivariate probit case. The outer product gradient form of LM test is not used so the proposed test is likely to perform reasonably well in small samples. The test is compared with two other existing tests.
110 - PublicationA simple artificial regression based LM test of asymmetry in the logit modelA simple and convenient artificial regression based LM tests of asymmetry in the logit model is derived. The test does not use the outer product gradient (OPG) form and in thus likely to have fairly good small sample properties.
85 - PublicationEducation, religion and the 'first destinations' of recent school-leavers in Northern IrelandThis paper uses data forma cross-section survey of recent school-leavers in Northern Ireland to examine the relationship between personal and social characteristics and first destinations approximately 6 to 9 months after leaving school. Six destinations are distinguished - employment, YTP, unemployment, higher education, further education and other activity. Particular attention is paid to the roles of religion, qualifications and subject choice. Nested logit models are estimated and subjected to a battery of mis-specification tests. The results suggest that, after controlling for a standard range of factors, religion and number of qualifications, but not subject choice, are significant in explaining post school destinations. In particular, Catholic males who leave from fifth and lower sixth form are significantly less likely to be employed and more likely to be on a YTP scheme.
92 - PublicationModelling winners and losers in contingent valuation of public goods : appropriate welfare measures and econometric analysis.Contingent Valuation is now the most widely used method for valuing non-marketed goods in cost benefit analysis. Yet, despite the fact that many externalities manifest themselves as costs to some and benefits to others, most studies restrict willingness to pay (WTP) to being non-negative. This paper explores appropriate welfare measures for assessing losses and gains and demonstrates how these can be elicited explicitly. Statistical / econometric methods are presented for modelling such responses. Median WTP is estimated non-parametrically. Grouped regression / Tobit and grouped regression / hurdle models are used to identify the determinants of WTP and to estimate mean WTP.
608 - PublicationTesting normality in bivariate probit models : a simple artificial regression based LM testA simple and convenient LM test of normality in the bivariate probit model is derived. The alternative hypothesis is based on a form of truncated Gram Charlier Type series. The LM test may be calculated as an artificial regression. However, the proposed artificial regression does not use the outer product gradient form. Thus it is likely to perform reasonably well in small samples.
158 - PublicationArtificial regression based LM test of mis-specification for ordered logit modelsLagrange Multiplier tests for omitted variables , heteroscedasticity, incorrect functional form and asymmetry in the ordered logit model may be readily calculated using an artificial regression. The proposed artificial regression is both convenient and likely to have better small sample properties than the more common outer product gradient based artificial regression.
114 - PublicationA simple LM test for zero correlation in the censored bivariate probit modelA simple artificial regression based on Lagrange Multiplier (LM) test for zero correlation in the censored bivariate probit model is derived. The outer product gradient form of the LM test is not used so the proposed test is likely to have good small sample properties.
118 - PublicationArtificial regression based LM tests of mis-specification for ordered probit modelsLagrange Multiplier (LM) tests for omitted variables, heteroscedasticity, incorrect functional form, and non-normality in the ordered probit model may be readily calculated using an artificial regression. The proposed artificial regression is both convenient and likely to have better small sample properties than the more common outer product gradient (OPG) form.
127