Phylogenetic

Comparative
Methods
Learning from trees

Luke J. Harmon
 5 × 1038
1
2
1

P = 0.94
 P
α
α = 0.05

pH = 0.5

H = 63

pH = 0.5 pH
pH ̸= 0.5

pH = 0.5

p = 0.5 n = 100
P = 0.003

pH = 0.5

pH > 0.5

P = 0.006

α
2

α
α = 0.05

3
L(H|D) = P r(D|H)
 pH

! "
n H
L(H|D) = P r(D|p) = p (1 − pH )n−H
H H

! "
100 63
L(H|D) = p (1 − pH )37
63 H

pH

L pH
pH
p̂H p̂H = 0.63

pH

pH

! "
n
ln L = ln + H ln pH + (n − H) ln (1 − pH )
H

d ln L H (n − H)
= −
dpH pH (1 − pH )

d ln L
dpH pH
p̂H

H
p̂H − n−H
1−p̂H = 0
H
p̂H = n−H
1−p̂H
H(1 − p̂H ) = p̂H (n − H)
H − H p̂H = np̂H − H p̂H
H = np̂H
p̂H = H/n
H/n = 63/100 = 0.63

â

âi

âi a
âi a
 p̂H = 0.63

pH = 0.5

pH

pH = 1/2

L1
∆ = 2 · ln = 2 · (ln L1 − ln L2 )
L2
 ∆ L2
L1

χ2

χ2
χ2

χ2
∆

pH = 0.5

# $
ln L1 = ln 100
63 + 63 · ln 0.5 + (100 − 63) · ln (1 − 0.5)
ln L1 = −5.92

# $
ln L2 = ln 100
63 + 63 · ln 0.63 + (100 − 63) · ln (1 − 0.63)
ln L2 = −2.50
∆ = 2 · (ln L2 − ln L1 )
∆ = 2 · (−2.50 − −5.92)
∆ = 6.84

χ2 P = 0.009

AIC = 2k − 2lnL
n/k > 40

2k(k + 1)
AICC = AIC +
n−k−1

AICc
AICc
∆AICc
∆AICc

∆AICci = AICci − AICcmin

AICci AICc AICcmin AICc

AIC ∆AICci
AICc
∆AICci
∆AICci

e−∆AICci /2
wi = % −∆AIC /2
ie
ci

wi
 AICc

AIC1 = 2k1 − 2lnL1 = 2 · 0 − 2 · −5.92

AIC1 = 11.8
AIC2 = 2k2 − 2lnL2 = 2 · 1 − 2 · −2.50
AIC2 = 7.0

n = 100

AICc1 = AIC1 + 2kn−k

1 (k1 +1)
1 −1
2·0(0+1)
AICc1 = 11.8 + 100−0−1
AICc1 = 11.8
AICc2 = AIC2 + 2kn−k
2 (k2 +1)
2 −1
2·1(1+1)
AICc2 = 7.0 + 100−1−1
AICc2 = 7.0

AIC

AICc

AICc

∆AICc1 = AICc1 − AICcmin

= 11.8 − 7.0
= 4.8

∆AICci
%
i e−∆i /2 = e−∆1 /2 + e−∆2 /2
= e−4.8/2 + e−0/2
= 1.09

p
p
p

P r(D|H) · P r(H)
P r(H|D) =
P r(D)

P r(H|D)
 P r(D|H)

P r(H)

P r(D)

&
P r(D) = P r(H|D)P r(H)dH
H

P r(D)

P r(H) = 1
P r(D|H) = P (D) P r(H|D) = 1

pH
pH f (pH ) = 1 pH
ph

P r(D|H) · P r(H) P (H|pH , N )f (pH )

P r(H|D) = = '1
P r(D)
0
P (H|pH , N )f (ph )dpH

P r(D|H)

#N $
H (1 − pH )
N −H
pH
P r(H|D) = ' 1 #NH$
H (1 − pH )
pH N −H dp
0 H H
 (N + 1)! H
P r(H|D) = p (1 − pH )N −H
H!(N − H)! H

pH

pH
 p0
i=1
p
p′ Q(p′ |p)

p p′
P (p′ )
Rprior =
P (p)

p p′

Q(p′ |p) = Q(p|p′ ) a2 = 1

Q(p′ |p)
Rproposal =
Q(p|p′ )

L(p′ |D) P (D|p′ )

Rlikelihood = =
L(p|D) P (D|p)

Raccept = Rprior · Rproposal · Rlikelihood

x
x < Raccept p ′ pi = p′
p pi = p
 pi i

U (0, 1) p
Q(p′ |p) U (p − ϵ, p + ϵ)
ϵ ≤ 0.01 p′ p
p

p′
ϵ = 0.01
U (0.59, 0.61) p′ = 0.595

U (0, 1) p
p′

P (p′ ) 1
Rprior = = =1
P (p) 1

Q(p′ |p) = Q(p|p′ )

Rproposal = 1
Rlikelihood p p′ p
p′

! " ! "
N H 100
P (D|p) = p (1 − p)N −H
= 0.66 3(1 − 0.6)100−63 = 0.068
H 63

! " ! "
N ′H 100
P (D|p ) =
′
p (1 − p )
′ N −H
= 0.5956 3(1 − 0.595)100−63 = 0.064
H 63
 P (D|p′ ) 0.064
Rlikelihood = = = 0.94
P (D|p) 0.068

Raccept = Rprior · Rproposal · Rlikelihood = 1 · 1 · 0.94 = 0.94

x = 0.34
x Raccept
p′

p = 0.60

lnL = −3 p
p = 0.63
p
 P r(D|H1 )
B12 =
P r(D|H2 )

P (D|H)

AIC
AIC

pH = 0.63

AICc

AICc

pH = 0.5 pH = 0.63

#100 $
P r(D|H1 ) = 63 0.50.63 (1 − 0.5)100−63
= 0.00270
' 1 #100 $ 63
P r(D|H2 ) = p=0$ 63
p (1 − p)100−63
#100
= 63 β(38, 64)
= 0.0099
0.0099
B12 = 0.00270
= 3.67

β(x, y)
AIC

AIC

AIC
AIC

AIC
 1

z̄

t z̄(t)

z̄(0)

σ2

σ2 t
 z̄(0) = 0 σ2 =
1

σ2 = 1
 σ2 = 1
t = 100
z̄(0) = 0
σ 2 t = 100

z̄(0) = 0 t = 100 σ2 = 1

σ2

σ2 = 1 σ2 = 5 σ 2 = 25

z̄(t)
 t = 100 σ2 σ2 = 1
2 2
σ = 10 σ = 25
 E[z̄(t)] = z̄(0)

z̄(t) ∼ N (z̄(0), σ 2 t)
E[z̄(t)] = z̄(0)
t
z̄(t)

z̄(0)
z̄(0)

0 t
t + ∆t

t + ∆t z̄(t + ∆t)
t z̄(t)
z̄(t) − z̄(0) z̄(t + ∆t) − z̄(t)
z̄(0)
z̄(t) ∼ N (z̄(0), σ 2 t) z̄(t)
z̄(0) σ2 t
σ2

z̄
 Ne 2

0
2
σm

E[z̄(t)] = z̄(0)

2 2
σB σB

2
σB

t σa2

σa2
σa2 σa2

1
E[σa2 (t + 1)] = (1 − )E[σa2 (t)] + σm
2
2Ne
 Ne
2
σm
1 2
(1 − 2Ne )E[σa (t)]

Ne

2
σm
2
σaStart
t

1 t 2
E[σa2 (t)] = (1 − 2
) [σaStart − 2Ne σm 2
] + 2Ne σm
2Ne

1 t
(1 − 2Ne )
t

lim E[σa2 (t)] = 2Ne σm

2
t→∞

2
t σB (t)
σa2

2 tσa2
σB (t) =
Ne

σa2

2 tσa2 2
t · 2Ne σm 2
σB (t) = = = 2tσm
Ne Ne
 2
σm

2
2
σa2 h2 =
σa
σw 2
σw

2
2Ne σm
h2 = 2
σw

2 h2 σ w
2
σm =
2Ne

h2 Ne 2
σw
σa2

2
σw

2 2 h2 σ w
2
t
σB (t) = 2σm t=
Ne

t
σ 2 = h2 σ w
2
/Ne 4

σs2

2 h2 σ W
2
σB =( + σs2 )t
Ne

2
σB ≈ σs2
 2
σE

2 2
σB ≈ σE

1/tij tij

E[z̄(t + 1)] = z̄(t) + h2 s

2 h2 σ w
2
σB =
Ne

E[z̄(t)] = t · (h2 s)
 2 h2 σ w
2
t
σB (t) =
Ne

z̄(0) = t·(h2 s)
 t1
t1
z̄(0)
2
σB

2
E[z̄(t)] ∼ N (z̄(0), σB t1 )

t1
t1 t2 t3

z̄(0)
2
σB x̄a
t 1 + t2

2
x̄a ∼ N [z̄(0), σB (t1 + t2 )]

t1 + t 3

2
x̄b ∼ N [z̄(0), σB (t1 + t3 )]

x̄a x̄b
 x̄a x̄b

x̄a = ∆x̄1 + ∆x̄2

x̄b = ∆x̄1 + ∆x̄3

∆x̄1 ∆x̄2 ∆x̄3

σ 2 t 1 σ 2 t2
σ 2 t3 x̄a x̄b

2
cov(x̄a , x̄b ) = var(∆x̄1 ) = σB t1

z̄(0)

! 2 " ! "
σ (t1 + t2 ) σ 2 t1 2 t 1 + t2 t1
= σ = σ2 C
σ 2 t1 σ 2 (t1 + t3 ) t1 t 1 + t3

n n×n
n

C(1, 2) C(2, 1) C
C
 C

R σi2
σij
n m
C n×n R m×m

m×1 a z̄1 (0) z̄2 (0)

z̄m (0) m×m R
m a m · (m + 1)/2 R
m · (m + 3)/2

R C

V=R⊗C

V n·m×n·m
 ! "
z̄1 (0)
a=
!z̄2 (0) "
σ12 σ12
R=
σ12 σ22

C = [t1 ]

! " ! 2 "
σ12 σ12 σ 1 t1 σ12 t1
V=R⊗C= ⊗ [t1 ] =
σ12 σ22 σ12 t1 σ22 t1

a
V

! 2 " ! "
σ1 σ12 t 1 + t2 t1
V=R⊗C= ⊗
⎡ 2 σ12 σ22 t1 t 1 + t3 ⎤
σ1 (t1 + t2 ) σ12 (t1 + t2 ) σ12 t1 σ12 t1
⎢σ12 (t1 + t2 ) σ22 (t1 + t2 ) σ12 t1 σ22 t1 ⎥
=⎢ ⎥
⎣ σ12 t1 σ12 t1 σ12 (t1 + t3 ) σ12 (t1 + t3 )⎦
σ12 t1 σ22 t1 σ12 (t1 + t3 ) σ22 (t1 + t3 )

a = [z̄1 (0), z̄1 (0), z̄2 (0), z̄2 (0)]

z̄(0)
 =0
z̄(0) ̸

x1 ∼ N (0, σ 2 t1 ) x2 ∼ N (0, σ 2 t2 ) x3 ∼ N (0, σ 2 t3 )

2
σ =1 x1 = −1.6 x2 = 0.1
x3 = −0.3

xa = +x1 +x2 = 0−1.6+0.1 = −1.5 xb = +x1 +x3 = 0−1.6+−0.3 = −1.9

xa xb
 Ne N

)
2
σw

x x2
2 2
σ = h /Ne
 z̄(0)

σ2

e ln
 n−1

i j
k
i j

cij = xi − xj
cij
vi + vj

cij xi − xj
sij = =
vi + vj vi + vj

σ2

(1/vi )xi + (1/vj )xj

xk =
1/v1 + 1/vj
 xk xi xj

k vk
vk + vi vj /(vi + vj )
xk

xi − xj
vi + vj
σ2

!
s2ij
σ̂P2 IC =
n−1

n−1
sij
(i, j)
n−1

σ̂P2 IC

2 · 0.09 = 0.18 √
2· 0.18 = 0.85
e2.68 = 2.3

! 2
e−1/2(x−z̄(0)1) (σ C) (x−z̄(0)1)
−1
2
L(x|z̄(0), σ , C) = "
(2π)n det(σ 2 C)

σ2 z̄(0) x n×1
n
C 1 n×1 (σ 2 C)−1
σ2 C

σ2 = 1 z̄(0) = 0

lnL(x|z̄(0), σ 2 , C) = −116.2

σ 2 = 0.09 z̄(0) = 4
σ 2 = 0.089 z̄(0) = 4.65
 2

2
σ̂M L = 0.08804487
z̄ˆ(0) = 4.640571 lnL = −78.04942

# $
314.6 −0.0026
H=
−0.0026 0.99

2
σ̂M L z̄ˆ(0)

0.06 − 0.11 2
σ̂M L 3.22 − 6.06 z̄ˆ(0)
 z̄ˆ(0) = (1C−1 1)−1 (1C−1 x)

2 (x − z̄ˆ(0)1)C−1 (x − z̄ˆ(0)1)
σ̂M L =
n

n C n×n
x
n×1 1 n×1
z̄ˆ(0) 2
σ̂M L

2
σ̂M L = 0.088
z̄ˆ(0) = 4.64

2 (x − z̄ˆ(0)1)C−1 (x − z̄ˆ(0)1)
σ̂REM L =
n−1

2 2
σ̂M L = 0.088 σ̂REM L = 0.090

σ̂P2 IC = σ̂REM
2
L

z̄ˆ(0) 3
 2
σ̂REM L = 0.090 1×1

2
σ̂REM L 0.05 − 0.13

C 60, 000 × 60, 000 = 3.6

C−1

σ2 z̄(0)

σ2 z̄(0)

Q(p′ |p)
wp
wp wp
Q(p′ |p) ∼ U (p − ,p + )
2 2

Rprior
p p

Rproposal
p p
Q(p′ |p) = Q(p|p′ )
Rproposal = 1
 Rlikelihood

L(p′ |D) P (D|p′ )

Rlikelihood = =
L(p|D) P (D|p)

Raccept

Raccept = Rlikelihood
x
x < Raccept

2
σ̂bayes = 0.10 0.066 − 0.15
z̄ˆ(0) = 3.5 2.3 − 5.3

σ2
 2
σ̂bayes
z̄ˆ(0)
 2
σ̂REM L
 x y x
y

x
y x y

y x

x y x
y

1
 b=1 n = 100

σx2 = σy2 = 2
1 σxy 2
σxy = 0
2 2
σxy = 0.8 σxy = −0.8

a r

! "
a = z̄1 (0) z̄2 (0) . . . z̄r (0)

R
 ⎡ ⎤
σ12 σ21 ... σn1
⎢ ⎥
⎢ σ21 σ22 ... ⎥
R=⎢
⎢
⎥
⎥
⎣ ⎦
2
σ1n ... ... σrn

σi2

σij = σji

R
C

V=R⊗C

V nr × nr V C
 !
e−1/2(xnr −D·a) (V) (xn r−D·a)
−1

L(xnr |a, R, C) = )
(2π)nm det(V)

D nr × r Dij (j − 1) · n < i ≤
j·n xnr
n n

! "
xnr = x11 x12 ... x1n x21 ... xnr

L(xnr |a, R, C)

â R̂

â = [(1C−1 1)−1 (1C−1 X)]!

(X − 1â)! C−1 (X − 1â)

R̂ =
n

X n ×r
 R x y R

* +
σx2 σxy
R=
σxy σy2

σxy

H1
H2 σxy

* 2 + * +
σ 0 σx2 σxy
RH1 = x RH2 =
0 σy2 σxy σy2

RH2
RH1
σx2 σy2

σx2 σy2
H1 H2

r = 0.27
P = 0.0001

* + * +
2.54 0.24 0.10
âH2 = R̂H2 =
4.64 0.10 0.09

lnL = −164.0
 * + * +
2.54 0.24 0
âH2 = R̂H2 =
4.64 0 0.09

lnL = −180.5
∆ = 33.0 P << 0.001
AICc

σx2 σy2 σxy z̄1 (0) z̄2 (0)

σx2 σy2 σxy

z̄1 (0) z̄1 (0)

Q(p′ |p)
Q(p′ |p) ∼
U (p − 0.1, p + 0.1)

Rprior

Rprior = 1
Rproposal

Q(p′ |p) = Q(p|p′ ) Rproposal = 1

Rlikelihood

L(p′ |D) P (D|p′ )

Rlikelihood = =
L(p|D) P (D|p)
 Raccept

Raccept = Rlikelihood
x
x < Raccept

σ̂x2 = 0.26 σ̂y2 = 0.10

σ̂xy = 0.11
σ̂xy
σ̂xy

x y

P << 0.0001

y x y
x=0
 y x

y = XD b + ϵ

y n×1 b
XD
y

⎡ ⎤
1 x1
⎢ 1 x2 ⎥
XD =⎢
⎣. . .
⎥
. . .⎦
1 xn

b 2×1
XD

y
ϵ

C
Ω

b̂ = (X!D Ω−1 X!D )−1 X!D Ω−1 y = (X!D C−1 X!D )−1 X!D C−1 y

b̂ z̄(0) b̂
y x
 x y
y x
x y
y x x
y

x y
x
 1

λ δ
κ
 λ λ

λ 0 ≤ λ ≤ 1
r

⎡ ⎤
σ12 σ12 ... σ1r
⎢σ21
⎢ σ22 ... σ2r ⎥
⎥
Co = ⎢ ⎥
⎣ ⎦
2
σr1 σr2 ... σr

⎡ ⎤
σ12 λ · σ12 ... λ · σ1r
⎢λ · σ21
⎢ σ22 ... λ · σ2r ⎥
⎥
Cλ = ⎢ ⎥
⎣ ⎦
2
λ · σr1 λ · σr2 ... σr

λ
λ

λ
λ>1

λ
λ λ=1

λ=0

α
 λ
λ

=0
λ̸

δ
δ
δ C
δ

⎡ ⎤
(σ12 )δ (σ12 )δ ... (σ1r )δ
⎢(σ21 )δ
⎢ (σ22 )δ ... (σ2r )δ ⎥
⎥
Cδ = ⎢ ⎥
⎣ ⎦
2 δ
(σr1 )δ (σr2 )δ ... (σr )

δ δ

δ>1

κ
κ
κ κ≥0

⎛ ⎞
1,1 +b1,2 ···+b1,d1
bk 1−2,1 +b1−2,2 ···+b1−2,d 1−r,1 +b1−r,2 ···+b1−r,d
k k
bk k k
... bk k k
1−2 1−r

⎜ bk2−1,1 +bk2−1,2 ···+bk2−1,d1−2 2,1 +b2,2 ···+b2,d2

bk k k
2−r,1 +b2−r,2 ···+b2−r,d
... bk k k
2−r ⎟
Co = ⎜
⎝
⎟
⎠
r−1,1 +br−1,2 ···+br−1,d
bk r−2,1 +br−2,2 ···+br−2,d r,1 +br,2 ···+br,dr
k k
bk k k
... bk k k
1−r 1−2

bx,y
x y dx,y
 x, y x

κ
κ = 0 κ
κ

AICc

λ
δ

AICc

AICc
2
σ = 0.088 θ = 4.64
σ2 = 0.085 θ = 4.64 λ = 1.0
σ2 = 0.063 θ = 4.60 δ = 1.5
σ2 = 0.170 θ = 4.64 κ = 0.66

AICc
AICc
 σ2

σ2

P = 0.42
 W =
251 P = 0.70

AICc

VH1 = σ 2 Ctree

C
σ12
σ22
C C1
σ12 C2
σ22
Ctree C1 C2
 VH2 = σ12 C1 + σ22 C2

H1
H2
H1
H2 σ12 = σ22

AICc

σ 2 = 0.088 z̄(0) = 4.64 lnL = −78.0 AICc = 160.4

σc2 σo2

σ̂c2 = 0.068 σ̂o2 = 0.01 z̄ˆ(0) = 4.51

lnL = −77.6 AICc = 162.3

σ 2 z̄(0) z̄(0)

z̄(0)
 z̄(0)

z̄(0) σ12 σ22

2 2 2
σdif f = σ1 − σ2
2
σdif f σ12
σ22
z̄0 σ 2 α θ
z̄ˆ0 = 4.60 σ̂ 2 = 0.10 α̂ = 0.0082 θ̂ = 4.60
AICc

σ2

σ 2 (t) = σ02 ebt

b
 µi (t) = z̄0
Vi (t) = σ02 e b −1 Vij (t) = σ02 e b −1
bTi bsij

z̄0 σ 2 b

z̄ˆ0 = 4.64 σ̂ 2 = 0.088

b̂ = −0.000001 lnL = −78.0 AICc = 162.6

σ2 α
 θ
z

2
W = e−γ(z−θ)

γ
γ2

W = 1 − γ(z − θ)2

W̄ = E[W ] = E[1 − γ(z − θ)2 ]

= E[1 − γ(z 2 − 2zθ + θ2 )]
= 1 − γ(E[z 2 ] − E[2zθ] + E[θ2 ])
= 1 − γ(z̄ 2 − Vz − 2z̄θ + θ2 )

z̄
 ∂ W̄
= −2γ z̄ + 2γθ = 2γ(θ − z̄)
∂ z̄

G ∂ W̄
∆z̄ =
W̄ ∂ z̄

G ∂ W̄ G
∆z̄ = = 2γ(θ − z̄)
W̄ ∂ z̄ 1 − γ(z̄ 2 − Vz − 2z̄θ + θ2 )

z̄ ′ = z̄ + 2Gγ(θ − z̄) + δ

z̄ z̄ ′
γ
θ δ

E[z̄ ′ ] = µ′z = µz + 2Gγ(θ − µz )

µz (0) =
z̄(0)

µz (t) = θ + e−2Gtγ (z̄(0) − θ)

Vz′ = E[z̄ ′2 ] + E[z̄ ′ ]2

Vz′ = E[(z̄ + 2Gγ(θ − z̄) + δ)2 ] − E[z̄ + 2Gγ(θ − z̄) + δ]2
Vz′ = G/n + (1 − 2Gγ)2 Vz
 Vz′ = G/n + (1 − 4Gγ)Vz

Vz (0) = 0

e−4Gtγ − 1
Vz (t) =
4nγ

z̄(0)
θ σ2
α z̄(0) θ α = 2G
σ 2 = G/n

ā′ = ā + 2Gγ(θ − ā) + δ

b̄′ = b̄ + 2Gγ(θ − b̄) + δ

µ′a = µa + 2Gγ(θ − µa )
µ′b = µb + 2Gγ(θ − µb )

µa (0) = ā0 µb (0) = b̄0

µ′a (t) = θ + e−2Gtγ (ā0 − θ)

µ′b (t) = θ + e−2Gtγ (b̄0 − θ)

a b
z
t1

E[ā0 ] = E[b̄0 ] = E[z̄(t1 )] = e−2Gt1 γ (z̄0 − θ)

 µ′a (t) = θ + e−2Gγ(t1 +t) (z̄0 − θ)
µ′b (t) = θ + e−2Gγ(t1 +t) (z̄0 − θ)

Va′ = E[ā′2 ] + E[ā′ ]2

Va′ = E[(ā + 2Gγ(θ − ā) + δ)2 ] + E[ā + 2Gγ(θ − ā) + δ]2
Va′ = G/n + (1 − 2Gγ)2 Va

Vb′ = E[b̄′2 ] + E[b̄′ ]2

Vb′ = G/n + (1 − 2Gγ)2 Vb

Va′ = G/n + (1 − 4Gγ)Va

Vb′ = G/n + (1 − 4Gγ)Vb

′
Vab = E[ā′ b̄′ ] + E[ā′ ]E[b̄′ ]
Vab = E[(ā + 2Gγ(θ − ā) + δ)(b̄ + 2Gγ(θ − b̄) + δ)] + E[ā
′

+2Gγ(θ − ā) + δ]E[ā + 2Gγ(θ − ā) + δ]

′
Vab = Vab (1 − 2Gγ)2

′
Vab = −4Vab Gγ

Va (0) = Va0 Vb (0) = Vb0 Vab (0) = Vab0

 1−e−4Gγt
Va (t) = 4nγ + Va0
Vb (t) = 1−e4nγ + Vb0
−4Gγt

Vab (t) = Vab0 e−4Gγt

z
t1

e−4Gγt1 − 1
Va0 = Vab0 = Vab0 = Vz (t1 ) =
4nγ

e−4Gγ(t1 +t) −1
Va (t) = 4nγ
−4Gγ(t1 +t)
Vb (t) = e 4nγ
−1
−4Gγ(t1 +t)
Vab (t) = e
−4Gγt
−e
4nγ

γ
γ

α = 2Gγ σ2 =
G/n

2
Va (t) = 2α
σ
(e−2α(t1 +t) − 1)
σ 2 −2α(t1 +t)
Vb (t) = 2α (e − 1)
σ 2 −2αt
Va b(t) = 2α e (1 − e−2αt1 )

µi (t) = θ + e−2GγTi (z̄0 − θ)

Ti
i

e−4GγTi − 1
Vi (t) =
4nγ

i j

e−4Gγ(Ti −sij ) − e−4GγTi

Vij (t) =
4nγ

Ti = Tj
i j
α σ2

µi (t) = θ + e−αTi (z̄0 − θ)

σ 2 −2αTi
Vi (t) = 2α e −1
σ 2 −2α(Ti −sij )
Vij (t) = 2α (e − e−2αTi )
 k
k=2
k k =2 k =3
k

q qij
i j
1

qij = qmn i ̸= j m ̸= n
 k=2 k=3
k=4
 ⎡ ⎤
−d1 q12 ... q1k
⎢ q21 −d2 ... q2k ⎥
⎢ ⎥
Q=⎢ ⎥
⎣ ⎦
qk1 qk2 ... −dk

di

k
'
d1 = q1i
i=2

k=2 q12 = q21

( )
−q q
Q=
q −q

k=3

P(t) = eQt

P (t)
e
2
e
P
pij i j
t
 e−t
k pij (t)
1/k π

* +
π = 1/k 1/k ... 1/k

* +
π = 1/2 1/2

qij = qji

k=2

k(k − 1)
p=
2
 * ,n−1 +
πSY M = π1 π2 ... 1− i=1 πi

n−1
n−1

k(k − 1)
p= +n−1
2
Q

⎡ ⎤⎡ ⎤
· r1 ... rn−1 π1 0 0 0
⎢ ⎥⎢
⎢ r1 · ... ⎥⎢ 0 π2 0 0⎥⎥
Q=⎢
⎢
⎥⎢
⎥⎣ ⎥
⎣ · rk(k−1)/2 ⎦ 0 0 0⎦
rn−1 ... rk(k−1)/2 · 0 0 0 πn

( )( ) ( ) ( )
· 1 0.75 0 · 0.25 −0.25 0.25
Q= = =
2 · 0 0.25 1.5 · 1.5 −1.5

k(k − 1)
n−1
 k =2
1/2
pstate0 = 0.5

P(t)

Q t
P(t)
q = 0.5
P(t)
t=3

( ) ( )
Qt −0.5 0.5 0.525 0.475
P(t) = e = exp( · 3) =
0.5 −0.5 0.475 0.525

u 0 ≤ u < 0.525
0.525 ≤ u < 1
3
 t
nchanges /t
t

P(t)
u
q = 0.5
t=3

1 1 9 −2·0.5·3
pii (t) = k + k−1
k e
−kqt
= 10 + 10 e = 0.145
1 1 −kqt 1 1 −2·0.5·3
pij (t) = k − ke 10 − 10 e = 0.095

P(t)

u
u
u = 0.475
u
 t = 3

q = 0.5

! " ! "
−0.5 0.5 0.525 0.475
P(t) = eQt = exp( · 3) =
0.5 −0.5 0.475 0.525

p12 (t)
t

L = 0.475
 q q = 1.0
q

q
q

q = 0.001850204 lnL = −80.487176

 q

Q(q ′ |q)
Q(q ′ |q) U (q − 0.1, q + 0.1)

Rprior
Rprior = 1
Rproposal
Rproposal
Rlikelihood

Raccept

Raccept = Rlikelihood
u
u < Raccept

q = 0.001980785 0.001174813 −
0.003012715

q
 q

p
1−p

n n0

Lgarbage = pn0 (1 − p)n−n0

p
p n0 /n

n = 258 n0 = 207 p =
n0 /n = 207/258 = 0.8023256
Lgarbage = pn0 (1 − p)n−n0 = 0.80232562 07(1 − 0.8023256)5 1 = 1.968142e − 56

lnLgarbage = n0 · ln(p) + (n − n0 ) · ln(1 − p) = 207 · ln(0.8023256) + 51 · ln(1 −

0.8023256) = −128.2677
lnL = −80.487176

! "
−q q
QE R =
q −q

! "
−q1 q1
QA SY =
q2 −q2

QE R

! "
−0.0019 0.0019
QE R =
0.0019 −0.0019
 ! "
−0.0016 0.0016
QA SY =
0.0038 −0.0038

AICc = 163.0 AICc = 162.8

q21
q12
 # #
LP (i) = ( P r(x|i, tL )LL (x)) · ( P r(x|i, tR )LR (x))
x∈k x∈k

i x k
x
i
 LL (x) LR (x)
tL tR

LP (i)

# #
LP (0) = ( P r(x|0, tL = 1.0)LL (x)) · ( P r(x|0, tR = 1.0)LR (x))
x∈k x∈k

P
tL = tR = 1.0

⎡ ⎤ ⎡ ⎤
−2 1 1 −2 1 1
Qt = ⎣ 1 −2 1 ⎦ · 1.0 = ⎣ 1 −2 1⎦
1 1 −2 1 1 −2

⎡ ⎤
0.37 0.32 0.32
P = eQt = ⎣0.32 0.37 0.32⎦
0.32 0.32 0.37

LL (0) = 1
LL (1) = LL (2) = 0 LR (1) = 1
LR (0) = LR (2) = 0

#
P r(x|0, tL = 1.0)LL (x) = 0.37 · 1 + 0.32 · 0 + 0.32 · 0 = 0.37
x∈k
 LP (0) = 0.37 · 0.32 = 0.12.

LP (1) = 0.32 · 0.37 = 0.12.

LL (x) LR (x)

#
L= πx Lroot (x)
x∈k
πx

πx = 1/k = 1/3

L = 1/3 · 0.00150 + 1/3 · 0.00151 + 1/3 · 0.00150 = 0.00150

t

Γ
 λ δ κ

δ κ

AICc

λ=0
δ > 0

AICc ∆AICc

AICc ∆AICc

λ
κ
δ
 AICc ∆AICc
λ
κ
δ

Q
r
rQ Q r

Q r

Q Q1 Q2 . . . Qn
 Q

y
x x
t y y x
 yi

yi

i pi (x)

t
Q
 ⎡ ⎤
−q1 − q2 q1 q2 0
⎢ q1 −q1 − q2 0 q2 ⎥
Q=⎢ ⎥
⎣ q2 0 −q1 − q2 q1 ⎦
0 q2 q1 −q1 − q2

⎡ ⎤
−q1 − q2 q1 q2 0
⎢ q1 −q1 − q3 0 q3 ⎥
Q=⎢ ⎥
⎣ q2 0 −q2 − q4 q4 ⎦
0 q3 q4 −q3 − q4

2
σ12
λ µ

t0

λ
µ

λ+µ
λ/(µ + λ) µ/(µ + λ)

N (t) t

N (t)(λ + µ)

r ϵ
 r =λ−µ

∆t

P rspeciation = N (t)λ∆t
P rextinction = N (t)µ∆t

t
N (t) N (t)
N (t)
N (t) ∆t

N (t + ∆t) = N (t) + N (t)λ∆t − N (t)µ∆t

N (t)
∆t ∆t

dN/dt = N (λ − µ)

N (0) = n0 0 n0

N (t) = n0 eλ−µ t = n0 ert

λ > µ r > 0

N (t)
N (t) pn (t) = P r[N (t) =
n] n≥0

n n t
pn (t)
 r=
λ−µ > 0 r=0 r<0
n0 = 1000
 p0 p∞

∆t ∆t

n=0
n=0
n = 1

p0 (t + ∆t) = p1 (t)µ∆t + p0 (t)

n = 1 n = 1
n=2

p1 (t + ∆t) = p1 (t)(1 − (λ + µ))∆t + p2 (t)2µ∆t

n>1 n
n−1 n n+1 n n n

pn (t + ∆t) = pn−1 (t)(n − 1)λ∆t + pn+1 (t)(n + 1)µ∆t + pn (t)(1 − n(λ + µ))∆t

pn (t) ∆t ∆t
n=0

dp0 (t)/dt = µp1 (t)

dp1 (t)/dt = 2µp2 (t) − (λ − µ)p1 (t)

n>1

dpn (t)/dt = (n − 1)λpn−1 (t) + (n + 1)µpn+1 (t) − n(λ + µ)pn (t)

pn (t) N (0) = n0
p0 (t) = 1 n = n0
 p0 (t)

p0 (t) = α0n

n≥1

min(n0 ,n) " #" #

! n0 n − 1 n0 −j n−j
pn (t) = α β [(1 − α)(1 − β)]j
j=1
j j−1

α β

ϵ(er t − 1)
α=
(er t − ϵ)
(er t − 1)
β=
(er t − ϵ)

α t
n0 = 1

p0 (t) = α

n≥1

pn (t) = (1 − α)(1 − β)β i−1

n0 t λ µ
 tc = 0
tstop
ti

Nalive ∗ (λ + µ) Nalive

tc + ti > tstop
λ/(λ+
µ) µ/(λ+µ)
ui [0, 1]
ui < λ/(λ + µ)

tstop
 n
a b
Ntotal Na + Nb Na
Nb
Ntotal Na + Nb
Ntotal = 10 1+9 2+8
3+7 4+6 5+5 6+4 7+3 8+2 9+1

1
p(Na | Ntotal ) =
Ntotal − 1

Ic

Ic

$
(NL − NR )
IC = allnodes
(N − 1)(N − 2)/2

N
 IC

IC = 1
IC

IC = 1 0 < IC < 1
IC = 0
 ln(n)
r̂ =
tstem

n tstem

r = λ−µ
 tcrown

ln(n) − ln(2)
r̂ =
tcrown

ϵ = µ/λ

log[n(1 − ϵ) + ϵ]
r̂ =
tstem

√
2
) (1−ϵ) n(nϵ2 −8ϵ+2nϵ+n)
ln[ n(1−ϵ
2 + 2ϵ + 2 ] − ln(2)
r̂ =
tcrown

ϵ=0 ϵ
ϵ
ϵ = 0.1 ϵ = 0.5 ϵ = 0.9
 r̂pb r̂ϵ=0.1 r̂ϵ=0.5 r̂ϵ=0.9

ϵ = 0.9

ϵ=0 ϵ = 0.9

N (0) = n0

N t tstem tcrown
λ µ
 r̂ = 0.27 ϵ̂ = 0

r ϵ

λprior

r ϵ

Q(p′ |p)
wp Q(p′ |p) U (p − wp /2, p +
wp /2)

p p′

′
λprior e−λprior p
Rprior = = eλprior (p−p )
′

λprior e−λprior p

p p′
Q(p′ |p) = Q(p|p′ ) Rproposal = 1
 Raccept

Raccept = Rprior · Rlikelihood

u
u < Raccept

λprior = 1 r ϵ r
ϵ

r ϵ
 Nn
Na Nb

Nn Na Nb
Na < N b Na + Nb = Nn
= Nb
Na ̸
2Na
P =
Nn − 1

Na = Nb P >1 P =1
Na = 81 − 46 = 35

2Na 2 · 35
P = = = 0.875
Nn − 1 81 − 1

N1
N0

N0
P =
Nn − 1

!
χ2 combined = −2 ln(Pi )

Pi i

χ2 combined
 2k

nf leshy ndry

P < 0.05
 χ2 combined = 72.8
2
χ P = 0.00027

P < 0.05 χ2 combined = 58.9

P = 0.0094

lnL = −215.6 lnL = −205.7 P = 0.000008

λ1 /λ0 = 1.39

Ic
Ic

Ic − n ∗ log(n) − n(γ − 1 − log(2))

Ic =
′

n
Ic = 1010 Ic = 3.57
′

P = 0.0004
 t1
t1

0 n

n n−1 t1
t2 t3 tn−1
t1 > t2 > · · · > tn−1 > 0

t1 − t2
n

t 1 t2
tn−1 n
λ µ

DN (t)

E(t)
 E(t)

DN (t) E(t)

DN (troot )
DN (t) E(t)
 DN (t)

E(t) DN (t)
E(t)
DN (t)
i
DN (t) = 1
E(t) = 0

tN
DN (t) E(t)
 dDN (t)
= −(λ + µ)DN (t) + 2λE(t)DN (t)
dt

−(λ+µ)DN (t)
2λE(t)DN (t)

dE(t)
= µ − (µ + λ)E(t) + λE(t)2
dt

µ −(µ +
λ)E(t)
λE(t)2 DN (t)

λ>µ

λ−µ
E(t) = 1 −
λ − (λ − µ)e(λ−µ)t

(λ − (λ − µ)e(λ−µ)tN )2
DN (t) = e−(λ−µ)(t−tN ) · DN (tN )
(λ − (λ − µ)e(λ−µ)t )2

tN
DN (t) E(t)
 DN ′ (t) = DN (t)DM (t)λ

" 2n−2
# (λ − (λ − µ)e(λ−µ)tk,t )2 $
L(t1 , t2 , . . . , tn ) = λn−1 e(λ−µ)(tk,b −tk,t ) ·
k=1
(λ − (λ − µ)e(λ−µ)tk,b )2

n
n

2n−2
k tk,b
tk,t

λ[1 − E(troot )]2

(n − 1)!
(n − 1)! n−1
n

" %2n−2 (λ−(λ−µ)e(λ−µ)tk,t )2 $

λn−2 e(λ−µ)(tk,b −tk,t ) ·
k=1 (λ−(λ−µ)e(λ−µ)tk,b )2
L(τ ) = (n − 1)!
[1 − E(troot )]2
 λ−µ
E(troot ) = 1 −
λ − (λ − µ)e(λ−µ)troot

λ µ µ
λ

n−2
λ=
sbranch

sbranch

n−1
!
sbranch = ti + tn−1
i=1

λ µ

λ = 0.46
µ = 0.20 lnLbd = 262.3
λ = 0.35 lnLpb = 260.4
AICbd = −520.6
AICpb = −518.8
∆ = 3.7
P = 0.054
λ = 0.48 µ = 0.23

b d
 f

DN (0) = 1 − f
E(0) = f

" 2n−2
# (f λ − (µ − λ(1 − f ))e(λ−µ)tk,t )2 $
L(t1 , t2 , . . . , tn ) = λn−1 e(λ−µ)(tk,b −tk,t ) ·
k=1
(f λ − (µ − λ(1 − f ))e(λ−µ)tk,b )2

" %2n−2 (f λ−(µ−λ(1−f ))e(λ−µ)tk,t )2 $

λn−2 e(λ−µ)(tk,b −tk,t ) ·
k=1 (f λ−(µ−λ(1−f ))e(λ−µ)tk,b )2
L(τ ) = (n − 1)!
[1 − E(troot )]2

λ−µ
E(troot ) = 1 −
λ − (λ − µ)e(λ−µ)troot
 Ltotal = Lphylogenetic · Ltaxonomic

Lphylogenetic Ltaxonomic

n−1
 AICC

λB µB
λA µA

λT µT
λA µA
λB µB

λA µA
λB µB

λT µT
λT = λA = λB µT = µA = µB
λT µT λliop µliop λother
µother

AICc
λT = 0.30
µT = 0.28
lambdaliop = 0.010
µliop = 0.007
lambdaother = 0.29
µother = 0.27

AICc

AICC
AICC AICC
r =λ−µ r
 E(t)

λ(t) µ(t)

!n−1
fn λ(ti )Ψ(si,1 , ti )Ψ(si,2 , ti )
L(t1 , t2 , . . . , tn−1 ) = (n + 1)! i=1
λ[1 − E(t1 )]2
 "t # " t " τ [λ(σ)−µ(σ)]dσ
[λ(u)−µ(u)]du e 0 λ(τ )dτ $−2
Ψ(s, t) = e s 1+ s
" s " τ [λ(σ)−µ(σ)]dσ
1
f + 0e 0 λ(τ )dτ

n
t1 , t2 , . . . , tn−1
λ(t) µ(t)
f
ti si,1 si,2

E(t) t
Ψ(s, t)
s t<s

ti n
n−1

(n + 1)!
λ(t) = λc µ(t) = µc

f =1

λ(t) = λ0 + αλ t

λ0
αλ
αλ > −λ0 /t1 αλ
αλ

µ(t) = µ0 + αµ t

αµ
αλ
αµ > −µ0 /t1

λ(t) = λ0 eβλ t

µ(t) = µ0 eβµ t

f = 440/471 = 0.934

λ = 0.05111267
 λ = 0.05111267
µ=0
λ0 = 0.035
αλ = 0.0011
µ=0
λ0 = 0.040
βλ = 0.016
µ=0
λ = 0.053
µ0 = 0
αµ = 0.000036
λ = 0.069
µ0 = 61.9
βµ = −111.0

Nt
λ(t) = λ0 (1 − )
K
 (n + 1)!

λ0 = 0.099 µ = 0 K = 979.9
 λ1

λ2
λ1 = 0.059 λ2 = 0.44 µ = 0.0
 q01
q10

λ0 µ0
λ1 µ1

n t k
n−k

ρ = k(q01 + λ0 + µ0 ) + (n − k)(q10 + λ1 + µ1 )

pq01 = (n · q01 )/ρ

pµ1 = [(n − k) · µ1 ]/ρ

 q01 = 110 = 0.05 λ0 = 0.2
λ1 = 0.8 µ0 = µ1 = 0.05
 ∆t

∆t → 0

p0 (t) p1 (t)

t + ∆t p0 (t + ∆t) p1 (t + ∆t)
t t + ∆t
p0 (t + ∆t) p1 (t + ∆t)
∆t

t+∆t

t + ∆t
 p0 (t + ∆t) = (1 − µ0 )∆t · [(1 − q01 ∆t)(1 − λ0 ∆t)p0 (t) + q01 ∆t(1 − λ0 ∆t)
p1 (t) + 2 · (1 − q01 ∆t)λ0 ∆t · E0 (t)p0 (t)]

[∆t]2 ∆t

p0 (t + ∆t) = [1 − (λ0 + 0 + q01 )∆t]p0 (t) + (q01 ∆t)p1 (t) + 2(λ0 ∆t)E0 (t)p0 (t)

p1 (t + ∆t) = [1 − (λ1 + 1 + q10 )∆t]p1 (t) + (q10 ∆t)p0 (t) + 2(λ1 ∆t)E1 (t)p1 (t)

p1 (t + ∆t)/[∆t] ∆t → 0

dp0
= −(λ0 + µ0 + q01 )p0 (t) + q01p1 (t) + 2λ0 E0 (t)p0 (t)
dt

dp1
= −(λ1 + µ1 + q10 )p1 (t) + q10p1 (t) + 2λ1 E1 (t)p1 (t)
dt

E0 (t) E1 (t)
t

dE0
= µ0 − (λ0 + µ0 + q01 )E0 (t) + q01 E1 (t) + λ0 [E0 (t)]2
dt
 dE1
= µ1 − (λ1 + µ1 + q10 )E1 (t) + q10 E0 (t) + λ1 [E1 (t)]2
dt

k
λ0 , λ 1 , . . . , λk µ0 , µ1 , . . . , µk
 q01 q10 λ0 λ1 µ0 µ1
q01 q10 λall µall
λ0 = λ1 = λall µ0 = µ1 = µall

AICc

= q10
q01 ̸

λ = 0.65
µ = 0.57

q01 = 0.16
q10 = 0.09
µ = 0.57

λ0 = 0.69
λ1 = 0.63
q01 = 0.17
q10 = 0.08
λ = 0.65

µ0 = 0.45
µ1 = 0.67
q01 = 0.22
q10 = 0.06
 λ0 = 0.49
λ1 = 0.79

µ0 = 0.20
µ1 = 0.84
q01 = 0.29
q10 = 0.05

q01 , q10 , λ0 , λ1 , µ0 , µ1

λpriori

Q(p′ |p)
wp Q(p′ |p) U (p − wp /2, p +
wp /2)

p p′

′
λpriori e−λpriori p
Rprior = = eλpriori (p−p )
′

−λpriori p
λpriori e

p p′
Q(p′ |p) = Q(p|p′ )
Rproposal = 1
 Raccept

Raccept = Rprior · Rlikelihood

u
u < Raccept

λ0 λ1 µ0
µ1