Latent trajectory models: an appetizer

1
Latent trajectory modelsan appetizer

• Tim Croudace

2
Enuresis repeated binary measure data 010101
3
Repeated binary outcome (night wetting
Y/N) Prevalence () at ages 4, 6, 8, 9, 11 and
15 n3272 (listwise deletioncomplete data on
all 6 occasions) NIW50 0.115 Night wetting
in past month at age 4 years NIW52
0.092 6 NIW54 0.059 8
NIW55 0.049 9 NIW57 0.050
11 NIW61 0.020 15
YProbwet last mnth
XTime/Age Yrs
4
NSHD Enuresis dataFrequencies of response
patterns

• RESPONSE PATTERNS
• No. Pattern No. Pattern No. Pattern
  No. Pattern
• 1 000000 2 100000 3 110000
  4 000001
• 5 011010 6 000100 7 111110
  8 011000
• 9 011100 10 111100 11 010000
  12 000010
• 13 011011 14 010010 15 000110
  16 001000
• 17 011110 18 111111 19 001010
  20 111010
• 21 001100 22 010001 23 001111
  24 101000
• 25 010111 26 101110 27 001110
  28 111000
• 29 010101 30 010100 31 100110
  32 011111
• 33 100100 34 110100 35 000011
  36 100101
• 37 110001 38 010011 39 001101
  40 100001
• 41 010110 42 110010 43 011101
  44 110110
• 45 100010

5
Boys
Girls
6
(No Transcript)
7
(No Transcript)
8
Latent Class Analysis Enuresis data 6 binary
(0/1)
Age 15 NIW61
Age 6 NIW52
Age 8 NIW54
Age 9 NIW55
Age 11 NIW59
Age 4 NIW50
u4
u2
u3
u1
u3
u1
u5
u6
u2
u4
y4
Latent classes Binary indicator prevalence
estimated through a logistic regression
intercept
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
Latent Class Growth Analysis Enuresis data 6
binary (0/1)
Age 15
Age 6
Age 8
Age 9
Age 11
Age 4
Binary outcome Modelled through Logistic
regression- Threshold fixed across occasions
u3
u1
u5
u6
u2
u4
y4
slope quadratic
intercept
1 1 1 1 1 1 1 4 6 8 9 11 15
16 36 64 121 225 Growth factors All
variances and covariances for growth factors0

13
(No Transcript)
14
(No Transcript)
15
(No Transcript)