The purpose of this session
• To introduce the practice dataset (MUSP)
• To examine distributions
• To perform a preliminary analysis: know
your data!
The MUSP Cohort
Mater-University Study of Pregnancy
• 8556 pregnant women recruited at first
antenatal visit 1981-1984 at the Mater
Mother’s Hospital, Brisbane
• Follow-up of mother-child sets
The MUSP cohort
• Follow-ups (mother and child) at delivery +
3 days, 6 months, 5 years, 14 years, 21
years (ongoing)
• Current cohort of approximately 5000
• Some repeated pregnancies, multiple births
Information collected
• Socio-demographic information, health behaviour
• Mental health, stressors, family functioning,
relationships
• Biological data: pregnancy, delivery history
• Physical, developmental assessment of child
• Child behaviour
• Health outcomes (mother and child)
Data structures
• Individual ID number (codea)
• Longitudinal record
codea a1 a2 a3…b1 b2 b3…c1 c2 c3….
• Restructure for ‘repeated measures’ analysis
Longitudinal file
Format:
codea a b c1 d1……. c2 d2……. c3 d3…….
1234 1 4 15.2 6….…. 16.6 9……. 18.6 11…….
repeated measures
Stacked file Time-constant
Format: variables
codea time a b c d
1234 1 1 4 15.2 6
1234 2 1 4 16.6 9 Time-dependent
1234 3 1 4 18.6 11 variables
Data Sets
Longitudinal file:
mps.SinglePregnancyMother
Used for examining change over time within
persons
Stacked file: mps.MentalHealth
Used for repeated measures analysis
The research questions
• Does mother’s mental health change over
time?
• What factors (time-constant, time-
dependent) predict mother’s mental health?
• What factors predict change in mother’s
mental health?
Mother’s mental health
• 14 ordinal items on a five point scale:
summed to make score 10 (good) – 50 (bad)
• Measured at six times so far:
FCV, + 3 days, + 6 months, + 5 years,
+ 14 years, + 21 years
• Missing values and attrition
Possible predictors
• Age at First Clinic Visit
• Health status: time-dependent
• …………
Descriptive analysis
• Missing values and attrition
• Distribution
• Change over time
• Correlations over time
• Determinants of mental health
Descriptive analysis
• Missing values and attrition
Attrition and Missing Values
• Attrition high; include only those present at
all phases FCV to +14 years
N =4470
• Missing values elsewhere on particular
items, comparatively infrequent.
Descriptive analysis
• Distribution (uses stacked file)
Descriptive analysis
• Change over time – grouped (uses stacked)
Summary statistics for mental health score (high=poor)
Variable Label N Mean Median Std Dev Minimum Maximum Skewness Kurtosis
amh FCV 4411 15.38 14.30 4.61 10.00 50.00 1.47 3.00
bmh + 3 days 4463 15.06 13.60 4.67 10.00 50.00 1.87 6.57
emh + 6 months 4456 14.81 13.60 4.67 10.00 50.00 1.68 4.15
hmh + 5 years 4455 16.19 15.00 5.12 10.00 44.30 1.35 2.46
kmh +14 years 4461 16.63 15.70 5.50 10.00 50.00 1.19 1.48
nmh +21 years 676 17.92 17.10 5.69 10.00 48.60 1.26 2.83
Summary statistics for log(mental health score-9)
Variable Label N Mean Median Std Dev Minimum Maximum Skewness Kurtosis
lamh FCV 4411 1.593 1.668 0.753 0.000 3.714 -0.278 -0.465
lbmh + 3 days 4463 1.515 1.526 0.790 0.000 3.714 -0.234 -0.613
lemh + 6 months 4456 1.454 1.526 0.803 0.000 3.714 -0.072 -0.778
lhmh + 5 years 4455 1.705 1.792 0.775 0.000 3.564 -0.397 -0.404
lkmh +14 years 4461 1.749 1.902 0.805 0.000 3.714 -0.419 -0.474
lnmh +21 years 676 1.966 2.092 0.714 0.000 3.679 -0.584 0.047
Poor
Mental
Health
Good
Mental
Health
Descriptive analysis
• Change over time – individual (uses
longitudinal)
Poor
Mental
Health Worsening
Improving
Good
Mental
Health
Poor
Mental
Health Worsening
Improving
Good
Mental
Health
Descriptive analysis
• Correlations over time
Correlations for mental health score
Pearson Correlation Coefficients
Number of Observations
FCV +3d +6m +5y +14y +21y
FCV 1.00000 0.54331 0.52288 0.42221 0.34201 0.31288
4411 4405 4397 4398 4402 665
+ 3 days 0.54331 1.00000 0.50318 0.42270 0.33176 0.33231
4405 4463 4449 4449 4455 674
+ 6 months 0.52288 0.50318 1.00000 0.49279 0.41533 0.36792
4397 4449 4456 4441 4447 675
+ 5 years 0.42221 0.42270 0.49279 1.00000 0.52577 0.37132
4398 4449 4441 4455 4446 673
+14 years 0.34201 0.33176 0.41533 0.52577 1.00000 0.51146
4402 4455 4447 4446 4461 676
+21 years 0.31288 0.33231 0.36792 0.37132 0.51146 1.00000
665 674 675 673 676 676
Correlations for log(mental health score-9)
Pearson Correlation Coefficients
Number of Observations
FCV +3d +6m +5y +14y +21y
FCV 1.00000 0.55385 0.52858 0.43313 0.35467 0.36037
4411 4405 4397 4398 4402 665
+ 3 days 0.55385 1.00000 0.55643 0.45615 0.36176 0.35962
4405 4463 4449 4449 4455 674
+ 6 months 0.52858 0.55643 1.00000 0.50648 0.43387 0.40551
4397 4449 4456 4441 4447 675
+ 5 years 0.43313 0.45615 0.50648 1.00000 0.53554 0.43299
4398 4449 4441 4455 4446 673
+14 years 0.35467 0.36176 0.43387 0.53554 1.00000 0.56196
4402 4455 4447 4446 4461 676
+21 years 0.36037 0.35962 0.40551 0.43299 0.56196 1.00000
665 674 675 673 676 676
= 0.456 = 0.684
Descriptive analysis
• Determinants of mental health
Marital Status at FCV
Cumulative Cumulative
Marital Status: a89 Frequency Percent Frequency Percent
Missing 33 0.74 33 0.74
SINGLE 348 7.79 381 8.52
DE FACTO 374 8.37 755 16.89
MARRIED 3638 81.39 4393 98.28
SEPARATED-DIVORCED 70 1.57 4463 99.84
WIDOWED 6 0.13 4469 99.98
PARTNER IN PRISON 1 0.02 4470 100.00
Combine “Other”
Poor
Mental
Health
Good
Mental
Health
Poor Excluding “missing” and “other”
Mental
Health
Means 2xSEM
Good
Mental
Health
Poor Single mothers poorer mental health at
Mental FCV, same at 6m, then widening gap
Health
Good
Mental
Health
Descriptive analysis
• Determinants of mental health
(time-dependent)
Health Status at 5, 14 and 21 years
Table of HStatus by time
HStatus(Health Status) time
Frequency + 5 +14 +21
Col Pct years years years Total
Missing 6 7 1 14
0.13 0.16 0.15
Excellent 1061 993 76 2130
23.74 22.21 11.23
Good 2603 2445 291 5339
58.23 54.70 42.98
Fair 687 872 234 1793
15.37 19.51 34.56
Poor 113 153 60 326
2.53 3.42 8.86
Very Poor 0 0 15 15
0.00 0.00 2.22
Total 4470 4470 677 9617
Frequency Missing = 3793
Poor Better physical health, better mental
Mental health, at all times
Health
Good
Mental
Health
Physical Health
Poor Better physical health, better mental
Mental health, at all times, association
Health stronger at 14 years
Good
Mental
Health
Physical Health
Time
Poor
Mental
Health
Poor Physical Health
Good
Mental
Health
Time
Excellent Physical Health
The MUSP dataset
• The subset available for all phases to 21 years (N = 667)
in a ‘stacked’ file: MentalHealth
• Outcome: Logged maternal mental health score
• Time-constant variables: At FCV: Marital status, Income,
Education, Age group, Number of previous pregnancies,
Country of birth, Time in Australia
• Time-dependent variables: Physical health status