Errata

thanks to
  Daniel F. Heitjan, University of Pennsylvania
  Keunbaik Lee, LSU

16_9:  can _be_ specified

20^12:  for _the_ binomial distribution

29^13-14:  allows _the_ logit

51^12:  can be used

51_6:  of _the_ normalizing constant

69^15:  Muller should have an umlaut

71_2:  Capitalize Polya

74^7:  posterior probability _was_ .11

77_9:  The last _row_ of Table 4.4

77_5:  The table says (-17, -7), not (-19, -5)

89^9: '...measurement at times $t_1...'

95^7:  x, _v_, and \psi

Page 111, line 6: Pr(y_mis|y_obs,\alpha_E) => Pr(y_mis|y_obs,r,\alpha_E)
               7: Pr(y_obs|\alpha_0) => Pr(y_obs|r,\alpha_0)

116^16:  refer _to_ the full-data

121_14:  in _a_ marginalized transition model

122^7:  i.e. _we_ specify

125_11:  \phi_{j,j+1} here and in legend to Table 6.1 disagrees with notation
as defined in eqn. (6.12)

137_15 and 137_17:  "t" should be math italic

150_17:  the _better_ fit

166^5:  illustrated _them_ using

Page 173, line -1: {\bar y}_J and {\bar y}_j should be boldfaced.

Page 174, line 1: {\bar y}_{j-1} should be boldfaced.

Page 185, line 4: p_k(y_j|y_1,...,y_{k-1}) => p_k(y_j|y_1,...,y_{j-1})

Page 193, line -15: Y_2|Y_1,S=1 => Y_2|Y_1,S=3

Page 195, line -9: k=1,2 => S=1,2

Page 196, line 13: as as => as

201^6:  (cf. the discussion

Page 207, line -10: In h_U(t_{ij}|b_i), b_i should be boldfaced.

219^1:  mainly _on_ the second

219^7:  so that _they_ are amenable

220_14:  is then based

222_18:  should be h(\xi_M,\Delta_0)

225_5:  part of the full-data _model_

234^5:  examines _the effects_ of

238^10:  we can make _alternative_ choices

238^12:  about _the_ treatment

244^8:   - E(Y-2 | Y_1, S\ge 2)

248_1:  in _the_ ET arm

249_2:  because we impose constraints on the full-data distribution and
on the missing data mechanism 

250_3:  -1 should also be in the set.