8.10 part iii: You can use the "It can be shown" statement on page 215 to justify the chi-square 
               part, and you can say that independence is assumed between terms in the ratio.

 --------------------------------------------------------------------------------------------------

9.6  The X matrix is n x 2, so X'X is 2 x 2. Invert that matrix, and the question calls for the
     diagonal entries in that matrix.

 --------------------------------------------------------------------------------------------------

9.15 parts a and b: An example very similar to this is examined on page 190. Be sure to at least briefly answer 
                    whether the various hypotheses are sensible. 

 --------------------------------
       an R coding hint:


# this is the rowbasis matrix from page 205
rowbasis <- matrix(c(1,1,0,0,
                     1,0,1,0,
                     1,0,0,1,
                    1,-1,-1,-1),nrow=4,ncol=4,byrow=T)
inrowbasis <- solve(rowbasis)
4*inrowbasis

 --------------------------------
       a SAS coding hint:
 * this shows how the solution is obtained for the example on page 205 ;

 proc iml ;
 * The rowbasis matrix from page 205 ;
 rowbasis = { 1  1  0  0  ,
 	      1  0  1  0  ,
              1  0  0  1  ,
	      1 -1 -1 -1   } ;

 invrowb = inv(rowbasis) ;
 invrowb2 = 4*invrowb ;
 print rowbasis ;
 print invrowb invrowb2 ;
 run ;


     part c: Use regression models with both dummy and deviation coding to generate the SS expressions.

 --------------------------------
       an R coding hint:

Moore1 <- read.table("c:/temp/moore.txt",header=F,skip=1)
names(Moore1) <- c("partner","status","conformity","fcategory","fscore")


 # defining deviation-coded variables
Moore1$stlowdev <- as.numeric(Moore1$status == "low") - as.numeric(Moore1$status == "high")
Moore1$flowdev <- as.numeric(Moore1$fcategory == "low") - as.numeric(Moore1$fcategory == "high")
Moore1$fmeddev <- as.numeric(Moore1$fcategory == "medium") - as.numeric(Moore1$fcategory == "high")

 # defining deviation-coded crossproduct variables used to test for interaction
Moore1$stLfLdev <- Moore1$stlowdev*Moore1$flowdev
Moore1$stLfMdev <- Moore1$stlowdev*Moore1$fmeddev


 # individual deviation-coded SSE terms from regression models 

SSEabc <- anova(lm(conformity ~ stlowdev +flowdev 
                                 +fmeddev +stLfLdev +stLfMdev, data=Moore1 ) )$"Sum Sq"[6]

SSEab <- anova(lm(conformity ~ stlowdev +flowdev 
                                 +fmeddev , data=Moore1 ) )$"Sum Sq"[4]

SSEac <- anova(lm(conformity ~ stlowdev +stLfLdev +stLfMdev, data=Moore1 ) )$"Sum Sq"[4]

SSEbc <- anova(lm(conformity ~ flowdev +fmeddev +stLfLdev +stLfMdev, data=Moore1 ) )$"Sum Sq"[5]

SSEa  <- anova(lm(conformity ~ stlowdev , data=Moore1 ) )$"Sum Sq"[2]

SSEb  <- anova(lm(conformity ~ flowdev +fmeddev , data=Moore1 ) )$"Sum Sq"[3]

SSEc <- anova(lm(conformity ~ stLfLdev +stLfMdev, data=Moore1 ) )$"Sum Sq"[3]

print(c(SSEabc,SSEab,SSEac,SSEbc,SSEa,SSEb,SSEc))

Note: SS(alpha | beta) = SSReg(alpha, beta) - SSReg(beta) = SSError(beta) - SSError(alpha, beta)
   this is the effect of the A factor added to a model already containing factor B,
   up above we have   SS(alpha | beta) = SSEb - SSEab = 212.21  (this is the Type II SS for factor A)

 the Anova function in the car() package calculates Type II SS

--------------------------------
      a SAS hint:

data moore ;
 infile 'c:\temp\moore.txt' firstobs = 2 ;
 input  partner status $ conformity fcategory $ fscore ;
 * here we define deviation-coded regressors, including interaction regressors ;
 statuslow = (status = 'low') ;
 if status = 'high' then statuslow = -1 ;
 fcategorylow = (fcategory = 'low') ;
 if fcategory = 'high' then fcategorylow = -1 ;
 fcategorymedium = (fcategory = 'medium') ;
 if fcategory = 'high' then fcategorymedium = -1 ;
 statlowfcatlow = statuslow*fcategorylow ;
 statlowfcatmed = statuslow*fcategorymedium ;
 title 'Moore and Krupat data, analysis from page 161';
 run ;

 proc reg data = moore ;
  modelSSabc : model conformity = statuslow fcategorylow fcategorymedium statlowfcatlow statlowfcatmed ;
  modelSSab : model conformity = statuslow fcategorylow fcategorymedium ;
  modelSSac : model conformity = statuslow statlowfcatlow statlowfcatmed ;
  modelSSbc : model conformity = fcategorylow fcategorymedium statlowfcatlow statlowfcatmed ;
  modelSSa : model conformity = statuslow ;
  modelSSb : model conformity = fcategorylow fcategorymedium ;
  run ;

 * proc glm does these calculations for us ;
  proc glm data = moore ;
  class status fcategory ;
  model conformity = status fcategory status*fcategory / ss1 ss2 ss3 ;
  run ;