Transcript Parametric Statistics

Assoc. Prof. Dr. Phongtape Wiwatanadate
LL.B., M.D., Ph.D.
Department of Community Medicine,
Faculty of Medicine,
Chiang Mai University
E-mail: [email protected]
Types of Measurement
Qualitative
Data
Quantitative Data
Qualitative Data
Categorical or Nominal Measurement
 Example: Blood Groups, Marital Status
 Quantitative comparison meaningless.
 Most of the times described as frequency or percentage.
 Outcomes with only two options (dead or alive, like or
dislike, male or female, etc.) called “Binomial
Measurement”.
Qualitative Data
Ordinal Measurement
Patient conditions (severe,
moderate, mild).
 Example:
 Individuals
can be ranked but
quantitative comparison is
impossible.
Quantitative Data
Data are always numeric and the
precision is unlimited.
 Mean and standard deviation can be
computed and meaningful.
 Quantitative comparison is possible
and meaningful.

Types of Variables
Dependent
Variable-Effect
Independent Variable-Cause
Types of Statistics
 Descriptive
Statistics
 mean, median, mode, standard
deviation
 Analytical
or Inferential Statistics
 Parametric
 Nonparametric
Parametric Statistics
Data to be analyzed must be quantitative and
dependent (except chi-square test).
 Data should be “Normally” distributed.
 The analyses are easy and contain high power
to detect the significance.


Even the distribution is somewhat deviated
from normal, the tests are still possible as long
as the sample size is big enough (i.e., > 30
samples).
Parametric Statistics
Examples
 Z-Test
 Student’s t-Test
 Analysis of Variance (One-Way, MultiWay)
 Chi-Square Test
 Regression
Nonparametric Statistics





Distribution-free statistics.
Analyses are impossible in case of too many parameters
in the study.
When sample size is big, calculations are complex and
tedious.
Power to detect significance is less than parametric.
Examples:





Sign Test
Wilcoxon Signed Rank Sum Test
Mann-Whitney U Test
Kruskal-Wallis Test
Friedman’s Test
Parametric Statistics
Mean Inference
One-Sample Tests
 Z-Test
(population variance
known)
 Student’s t-Test (population
variance unknown)
Parametric Statistics
Mean Inference (cont’d)
 Two-Sample Tests
 Paired
t-Test
 t-Test for Independent Samples with
Equal Variances
 t-Test for Independent Samples with
Unequal Variances
Parametric Statistics
Variance Inference
 One-Sample 2 Test
 Two-Sample F Test for Equality of Two
Variances
 Multisample:
 One-Way Analysis of Variance [variances
must be equal (Levene’s Test)]
 Multi-Way Analysis of Variance
Parametric Statistics
Categorical Data Inference
 Chi-Square Test
 Invalid if 20% of total cells have
expected values < 5.
 Invalid if expected values < 1.
 Fisher’s Exact Test
 Provides the exact result.
 Calculations are tedious and require high
performance computer.
Parametric Statistics
Regression
Linear Regression
Non-linear Regression
Multiple Regression
Logistic Regression
Poisson Regression