Sebastian Weirich
2013-May-02 14:27 UTC
[R] Package survey: singularities in linear regression models
Hello, I want to specify a linear regression model in which the metric outcome is predicted by two factors and their interaction. glm() computes effects for each factor level and the levels of the interaction. In the case of singularities glm() displays "NA" for the corresponding coefficients. However, svyglm() aborts with an error message. Is there a possibility that svyglm() provides output for coefficients without singularities like glm()? Thank you very much, Sebastian Weirich -- Sebastian Weirich, Dipl.-Psych. Institut zur Qualit?tsentwicklung im Bildungswesen Humboldt-Universit?t zu Berlin Sitz: Hannoversche Stra?e 19, 10115 Berlin Postadresse: Unter den Linden 6, 10099 Berlin Tel: +49-(0)30-2093-46512
Thomas Lumley
2013-May-02 20:02 UTC
[R] Package survey: singularities in linear regression models
On Fri, May 3, 2013 at 2:27 AM, Sebastian Weirich < sebastian.weirich@iqb.hu-berlin.de> wrote:> Hello, > > I want to specify a linear regression model in which the metric outcome is > predicted by two factors and their interaction. glm() computes effects for > each factor level and the levels of the interaction. In the case of > singularities glm() displays "NA" for the corresponding coefficients. > However, svyglm() aborts with an error message. Is there a possibility that > svyglm() provides output for coefficients without singularities like glm()? > >It's not true that svyglm() aborts with an error message whenever there are singularities, eg> svyglm(enroll~stype+I(stype),design=dclus1)1 - level Cluster Sampling design With (15) clusters. svydesign(id = ~dnum, weights = ~pw, data = apiclus1, fpc = ~fpc) Call: svyglm(formula = enroll ~ stype + I(stype), design = dclus1) Coefficients: (Intercept) stypeH stypeM I(stype)H I(stype)M 432.9 697.4 464.9 NA NA Degrees of Freedom: 182 Total (i.e. Null); 12 Residual Null Deviance: 24830000 Residual Deviance: 15120000 AIC: 2599 So, perhaps you could show us what you actually did, and what actually happened, as the posting guidelines request. -thomas -- Thomas Lumley Professor of Biostatistics University of Auckland [[alternative HTML version deleted]]