BIOSTATISTICS

Descriptive analysis
NEILY ZAKIYAH, PHD., APT
Descriptive statistics
To describe and summarize data

Small dataset: look at the raw data

Larger datasets: try to summarize the data with numbers and figures
Descriptive statistics
§ Type of variable
§ Distribution of variable
§ Summary statistics
§ Figures
Variable Interval:
Temperature, calendar
Classification of variables Continuous:
years, intelligence scale

blood pressure, height,

weight
Ratio:
Quantitative (numerical) Age, income earned per
Discrete: month
number of children,
number of asthma
attacks/year
Variable

Ordinal:
better/same/worse,
dyspnea degree 2,3,4
Qualitative
(categorical)
Nominal: Note: Some ambiguities in classifying
a type of variable may arise in some
gender, yes/no, blood
cases.
group
Distribution of variable
Normal and skewed distribution
Measurement of
central tendency:
Normal distribution:
mean
Skewed distribution:
median
Differences between groups
Purpose: to make statements about a population in a study

For example: all patients with antidepressants in a hospital

◦ Problem: you cannot measure everybody and everything
◦ Solution: take a random sample
Differences between groups:
Statistical approach
Ø Concept
Comparing two groups:
H0 : no difference between groups
H1 : difference between groups
P value: chance to observe H1 if H0 is true
Ø Testing
P < 0.05 chance to observe H1 if H0 is true < 0.05 (5%)
95% CI 95% chance the real difference between 2 groups falls within the interval
Differences between groups:
Statistical approach
Ø Testing:
Choice for the type of the test depends upon:
1.(In)dependence of observation

Independent, two different groups Dependent (paired or matched)

(Comparing two population means (Every observation in one sample is linked to an
based on two independent samples) observation in another sample)
e.g. same subject measure before and after
certain tretment

2.Type of variable and distribution of variable

Which test for which aim?
Non-parametric Purpose Parametric
Independent observations (unpaired)
Chi-square Categorical/ nominal data Fisher’s exact
(small and large sample) Chi-square
The Mann–Whitney U test Numerical/ continuous , ordinal data Independent t-test
/ Wilcoxon rank-sum test
Kruskal-Wallis Numerical/ continuous, ordinal data, test One-way analysis of
differences between >2 groups variance (ANOVA)
Paired/matched observations
McNemar’s Categorical/ nominal data (distribution free)
Wilcoxon signed rank test Numerical/ continuous, ordinal data Paired t-test
Friedman Numerical/ continuous, ordinal data, test Two-way analysis of
differences between >2 groups variance (ANOVA)
Reasons to choose non-parametric tests:
• Data with skewed distribution, meaning that the data is
better represented by median instead of mean.
• Data with very small sample size
• Ordinal data, ranked data and outliers
Choice of statistical test for independent observations

aIf data are censored. b The Kruskal-Wallis test is used for comparing
ordinal or non-Normal variables for more than two groups, and is a
generalisation of the Mann-Whitney U test. c Analysis of variance is a
general technique, and one version (one way analysis of variance) is
used to compare Normally distributed variables for more than two
groups, and is the parametric equivalent of the Kruskal-Wallistest. d If
the outcome variable is the dependent variable, then provided the
residuals (the differences between the observed values and the
predicted responses from regression) are plausibly Normally distributed,
then the distribution of the independent variable is not
important. e There are a number of more advanced techniques, such as
Poisson regression, for dealing with these situations. However, they
require certain assumptions and it is often easier to either dichotomise
the outcome variable or treat it as continuous. © MJ Campbell 2016, S
Shantikumar 2016
Choice of statistical test for paired or matched observations

© MJ Campbell 2016, S Shantikumar 2016

 Example:
Influenza vaccination trial (observed and expected numbers)

Effectiveness of influenza vaccination is assessed in a study, where 150 patients aged 65 years old
or older are getting the influenza vaccination, and another 150 patients in the same age group
are getting placebo. The study shows that the rate of influenza in the patients group with
vaccination is lower than in the patients group without vaccination. How to determine whether
the difference is statistically significant?

u Chi square is used to estimate whether the difference in influenza rate among two patients
group is statistically significant.
c2 = å ( O - E)2 / E df = (r-1) (c-1)

O: observed
E: expected
df: degree of freedom
Example Chi-square (X2)
Influenza vaccination trial in SPSS
Chi-square – 2x2 table – via Crosstabs (Analyze – Descriptive Statistics– Crosstabs)

P-value

Observed chi-square
 Example:
In a study, the anxiety scores are observed from 50 patients aged 10 - 15 years old with divorced
parents and 50 patients with non-divorced parents. The results shows that the average anxiety
score of patients with divorced parents is higher than patients with non-divorced parents. How
to determine whether the difference in average anxiety score is statistically significant?
Normality assumption is used in the anxiety score

u Independent t-test is used to estimate whether the difference in average anxiety score among
two patients group is statistically significant.
Example independent t-test
Anxiety score in SPSS
(Analyze – compare means – independent samples t-test)
 Example:

In a population based cohort study, Apgar scores at five and 10 minutes in infants born less than
37 completed weeks are observed. The Apgar score has been used worldwide as an index of
early neonatal condition. In its descriptive analysis, the Apgar score in five minutes is observed to
be different from the Apgar score in 10 minutes. How to determine whether the difference is
statistically significant?
Non-normality assumption is used in Apgar score

u The observation is in one group (paired)

u Wilcoxon signed rank test is used to estimate whether the difference in Apgar score in five
and 10 minutes measurement is statistically significant.
Exercise

In a retrospective cohort study, 3431 inhabitants of city A and city B who had histamine
challenge test data were assessed from year 1964 – 1972, for approximate 30 years. The
objective of the study was to assess association between histamine airway hyper-responsiveness
and mortality from chronic obstructive pulmonary disease.
Baseline characteristics of the study population on gender, age, height, smoking habits, and
several respiratory symptoms were collected. Descriptive statistics were conducted to assess
whether there were statistically significant differences in baseline characteristics between men
and women participants.
Please determine which test is appropriate to assess each characteristic based on the following
information (in the Table)!
Exercise
Characteristics of the participants of the study, for men and women separately

Men Women p-value Test

N(%) 1858 (54.1%) 1573 (45.8%)
Chi square
• City A 1168 (52.8%) 1044 (47.2%) 0.035
• City B 690 (56.6%) 529 (43.4%)
Age (years), median (25th-75th perc) 40.0 (30.0 – 50.0) 41.0 (31.0 – 50.0) 0.337 M-W

Height (cm), mean (sd) 177.1 (6.7) 163.9 (5.9) 0.000 Independent t-test

Smoking habits
• Never smokers, N(%) 252 ( 24.2%) 791 (75.8%) Chi square
• Ex-smokers, N(%) 444 (68.2%) 207 (31.8%) 0.000
• Current smokers, N(%) 1017 (64.0%) 572 (36.0%)
FEV1 (cl), mean (sd) 350.1 (79.7) 268.1.0 (53.0) 0.000 Independent t-test

Weight (kg), mean (sd) 79.5 (10.1) 68.9 (11.0) 0.004 Independent t-test

Peripheral blood eosinophil (cell/µl)

M-W
Median (25th-75th perc) 9 (5-16) 8 (4-14) 0.001