s-news
[Top] [All Lists]

drop1 v. ssType=3

To: <s-news@wubios.wustl.edu>
Subject: drop1 v. ssType=3
From: "Eric Turkheimer" <ent3c@virginia.edu>
Date: Tue, 26 Nov 2002 17:21:57 -0500
Importance: Normal
Organization: University of Virginia
Reply-to: <ent3c@virginia.edu>

Can someone explain to me the following behavior of ssType=3?  My understanding of Type3 SS is that it gives the reduction in SSE produced when each term is added to all the others, ie, like drop1.

 

#Define linear model

     worthlm4<-lm(slfwth1y ~  cdisum1y+ygesy1c , data="">

#Drop one term at a time

     drop1(worthlm4)

Single term deletions

 

Model:

slfwth1y ~ cdisum1y + ygesy1c

         Df Sum of Sq      RSS       Cp

  <none>              814.4197 848.5911

cdisum1y  1  103.5157 917.9354 940.7164

 ygesy1c  1    7.8926 822.3123 845.0932

 

#Shouldn't Type 3 SS give the same thing?

     anova(worthlm4,ssType=3)

Analysis of Variance Table

 

Response: slfwth1y

 

Type III Sum of Squares

           Df Sum of Sq  Mean Sq       F Value Pr(F)

 cdisum1y   1    0.0000   0.0000 0.000000e+000     1

  ygesy1c   1  158.2564 158.2564 1.063195e+034     0

Residuals 143    0.0000   0.0000          

 

Eric Turkheimer, PhD

Department of Psychology

University of Virginia

PO Box 400400

Charlottesville, VA  22904-4400

 

434-982-4732

434-982-4766 (FAX)

 

<Prev in Thread] Current Thread [Next in Thread>
  • drop1 v. ssType=3, Eric Turkheimer <=