Hi, I have the problem that for the step procedure stops due to missing values. There are no options in Step or stepAIC to handle missing values. Is there any way to run stepwise modelselection in R in an automated way in this case? Here is the last step before it stops. Hope someone knows. Best regards, Andreas Step: AIC= 1999.16 EF ~ SF120_KS + SF120_PS + HADA0 + SOZU0 + LVEDD + logPROBNP + ALTER + SD0_01 + ASE_UK + DS140POS + RSQSICH0 + SD0_01:ASE_UK + SD0_01:DS140POS + SD0_01:RSQSICH0 + ASE_UK:DS140POS + ASE_UK:RSQSICH0 + DS140POS:RSQSICH0 + SD0_01:ASE_UK:RSQSICH0 + SD0_01:DS140POS:RSQSICH0 + ASE_UK:DS140POS:RSQSICH0 Df Sum of Sq RSS AIC - SOZU0 1 3.0 25356.0 1997.2 - HADA0 1 7.6 25360.6 1997.3 - ALTER 1 13.0 25365.9 1997.4 - SF120_PS 1 14.7 25367.6 1997.5 - ASE_UK:DS140POS:RSQSICH0 1 20.1 25373.1 1997.6 - SD0_01:DS140POS:RSQSICH0 1 44.8 25397.7 1998.0 - SD0_01:ASE_UK:RSQSICH0 1 54.4 25407.4 1998.2 <none> 25352.9 1999.2 - LVEDD 1 382.2 25735.1 2004.6 - SF120_KS 1 476.4 25829.3 2006.4 - logPROBNP 1 891.9 26244.9 2014.4 Error in step(mod2, direction = "back") : number of rows in use has changed: remove missing values?
On Fri, 7 Oct 2005, Andreas Cordes wrote:> I have the problem that for the step procedure stops due to missing > values. There are no options in Step or stepAIC to handle missing > values. Is there any way to run stepwise modelselection in R in an > automated way in this case?Try the hint it gives you, or see the help page (which covers this in a warning with an explanation). [...]> Error in step(mod2, direction = "back") : > number of rows in use has changed: remove missing values?-- Brian D. Ripley, ripley at stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UK Fax: +44 1865 272595
At 11:11 7/10/2005, you wrote:>Hi, >I have the problem that for the step procedure stops due to missing >values. There are no options in Step or stepAIC to handle missing >values. Is there any way to run stepwise modelselection in R in an >automated way in this case? > >Here is the last step before it stops. Hope someone knows. Best regards, >Andreas > >Step: AIC= 1999.16 > EF ~ SF120_KS + SF120_PS + HADA0 + SOZU0 + LVEDD + logPROBNP + > ALTER + SD0_01 + ASE_UK + DS140POS + RSQSICH0 + SD0_01:ASE_UK + > SD0_01:DS140POS + SD0_01:RSQSICH0 + ASE_UK:DS140POS + >ASE_UK:RSQSICH0 + > DS140POS:RSQSICH0 + SD0_01:ASE_UK:RSQSICH0 + >SD0_01:DS140POS:RSQSICH0 + > ASE_UK:DS140POS:RSQSICH0 > > Df Sum of Sq RSS AIC >- SOZU0 1 3.0 25356.0 1997.2 >- HADA0 1 7.6 25360.6 1997.3 >- ALTER 1 13.0 25365.9 1997.4 >- SF120_PS 1 14.7 25367.6 1997.5 >- ASE_UK:DS140POS:RSQSICH0 1 20.1 25373.1 1997.6 >- SD0_01:DS140POS:RSQSICH0 1 44.8 25397.7 1998.0 >- SD0_01:ASE_UK:RSQSICH0 1 54.4 25407.4 1998.2 ><none> 25352.9 1999.2 >- LVEDD 1 382.2 25735.1 2004.6 >- SF120_KS 1 476.4 25829.3 2006.4 >- logPROBNP 1 891.9 26244.9 2014.4 >Error in step(mod2, direction = "back") : > number of rows in use has changed: remove missing values?Andreas Try data<-na.omit(original database) before you run step() or stepAIC() Bernardo Rangel Tura, MD, MSc National Institute of Cardiology Laranjeiras Rio de Janeiro Brazil --