Class Notes
(811,169)

Canada
(494,539)

University of Toronto Scarborough
(30,852)

Health Studies
(1,162)

HLTC15
(5)

Suzanne Sicchia
(5)

Lecture 7

# HLTC15 Lecture 7

by
OneClass7148

Unlock Document

University of Toronto Scarborough

Health Studies

HLTC15

Suzanne Sicchia

Fall

Description

HLTC15 –Introduction to Quantitative and Qualitative Health Research Methodologies
Lecture 7: Quantitative Research Design and Methods I–Sampling Designs (Chapter 9 & 11)
Tuesday, October 23, 2012
Scientific Method and the Process of Quantitative Research
1. Theory
2. Hypothesis (Ho)
3. Research Design
4. Devise Measures of Concepts
5. Select Research Site(s)
6. Select Subjects/ Respondents
7. Collect Data
8. Analyze Data
9. Write-up Findings and Conclusions
10. Disseminate
Common Characteristics of Positivist (Quantitative) Research
Interest in cause and effect relationships
Emphasis is on “objectivity” (scientific knowledge) over other
forms of “subjective” knowledge
Validity, reliability, generalizability, and replicability
Control of bias in scientific knowledge production
Reductionist: reduces social situations to smaller parts (often just two variables) for study
Statistics
The mathematics of the collection, organization, and interpretation of numerical data, especially the analysis of
population characteristics by inference from sampling
Three broad categories of statistical tests:
1. Means: Test for a mean; Difference of two means (independent samples); Difference of two means (paired tests)
2. Proportions: Tests for proportion; Differences of two proportions
3. Relationships: Chi-square test for independent, Regression Analysis
**Choosing the right statistics depends on the data, samples, and purpose of the study**
Rational and Some Basic Terminology
Why Sample?
o To collect data from a sub-set of a population so that valid inferences can be drawn about the whole
(generalizability, predictability, etc.)
Terminology:
o Target Population: the population of interest
o Sampling Unit or Element: member of the target population
o Sampling Frame: list of units in the target population
o Probability Sampling: where each population element has a specifiable chance of selection
o Non-probability sampling: use of judgement criteria
Thinking about Sample Design
Who do you want to generalize to? This is the theoretical population
What population can you get access to? This is the study population
How can you get access to them? Refers to the sampling frame
Who is in your study? The sample
Non-Probability Sample Designs (no random selection, typical in qualitative studies)
1. Convenience Sampling – Based on the ease of selection (e.g. patients attending a clinic)
2. Volunteers – Advertising for sample members, used in both qualitative and experimental research
3. Snowball Sampling – starting with an initial sample, and increasing size via referral by sample members 4. Purposive Sampling – selecting all sample elements according to explicit judgement criteria
5. Theoretical Sampling – qualitative approach in which sample selection is guided by the requirements of generating and
testing empirical hypotheses
Probability Sample Designs (random selection)
1. Simple Random Design
2. Systematic Sample Design
3. Stratified Sample Design
4. (Multi-Stage) Cluster Sample Design
1. Simple Random Sample (SRS)
All units in a given population have an equal probability of being included in the sample
Example: 30 students in a classroom and you want a SRS of 10 students:
1) Assign a # to all students
2) Randomly choose 10 :
Like a lottery
And if the same # comes up more than once, it is dismissed and another # is pulled
2. Systematic Sampling
Sample is selected directly from the sampling frame at some
interval after a random starting point has been chosen
o Example: Start with the 2nd person and then include
every third person until after that until you reach the
desired sample size (n)
3. Stratified Sample
Group population according to some characteristic (e.g., sex)
Each group is called a strata
Obtain a (proportional) simple random sample from each group
Proportionality ensures the sample is distributed in the same way as
the population
4. Cluster/ Multi-Cluster Samples
Not to be confused with “stratified” sampling
Obtained by selecting all individuals within a randomly selected
collection or group of individuals (clusters)
Works best when clusters are diverse (saves money)
Example: 5 Streets in a particular neighbourhood (five clusters)
Assign a # to each cluster, then use SRS to choose ONE cluster (or
street); interview every individual living on that street
Multi-Cluster Sample: when a cluster is sub-sampled one or more
times
Three Sources of Bias in Sample Design
1) Failure to use a random method to choose the sample (where every element or individual has the same chance of being
selected—like #s in a lottery)
2) When the sampling frame or list of potential subjects is inaccurate or excludes some cases the sample derived from it
may not represent the population (even with randomization)
3) Non-response: when some people refuse to participate – non-responders may be different from those who agree to
participate in some significant way Sampling Error (Standard Error)
The amount of inaccuracy in an estimation of some value caused by the sample
(a portion of the population) rather than the whole population
Low sampling error means we have relatively less variability or range in the
sampling distribution
The greater the sample size, the smaller the standard error (because your
sample is closer to the actual population)
Non-Sampling Error
Coverage Error: Sources
o Frame – how accurate is the sampling frame?
o Response – how high is the response rate?
o Response – how representative are responders?
Measurement Error: Sources
o Respondent – how accurate and reliable are they?
o Instrument – how valid and reliable is it?
o Recorder – how much influence on the data collected?
Chapter 11: Basic Statistical Methods for Health Data Analysis
Overview of Statistical Methods
Statistics: The study of the collection, organization, analysis, interpretation, and presentation of data
Two branches of statistical techniques involved in addressing research questions:
1. Descriptive Statistics: Summarize data collected using tables, graphs, or statistical measures such as the mean
and correlation coefficient (e.g., Pearson’s r)
2. Inferential Statistics: Help us determine whether or not we can generalize descriptive statistics for a random
sample to a wider population
Descriptive vs. Inferential Statistics
Statistics
1. Descriptive Statistics Collecting, Organizing, Summarizing, Presenting Data
2. Inferential statistics Making inferences, Hypothesis Testing, Relationships, Predictions
Statistics Descriptive Statistics
1. Graphs + Tables
Pie charts
Bar graphs
Histograms
Frequency (f) Tables
2. Numerical Measures
Measures of Central Tendency (Mean, Median, Mode)
Measures of Dispersion (Univariate): Range (IQR), Standard Deviation
Types of Measurements and their Function
Type Description Example: Measuring the health status of
patients
Values are distinct categories Classifying patients according to the organ of
Cannot be quantified or rank ordered the body affected by their disease (e.g.,
Nominal
(e.g., gender, marital status, religion) nervous system, respiratory system)
Only allow for qualitative classification
Values are distinguishable Classifying patients according to whether they
Ordinal Can be rank ordered (x is > or < than y) rate themselves as Very Unhealthy, Unhealthy,
But intervals between scale points are uneven, and Healthy, or Very Healthy
there is no meaningful zero point
Values are often ranked (e.g. 1 , 2 , 3 , ...) Values are distinguishable, continuous data Counting the number of times in the previous
Interal/ Variables can be rank ordered year a patient has consulted a doctor or other
ratio Distance between categories is equal health professional
Ratio: have an absolute zero point, ie., zero is no
arbitrary (e.g. weight, length, time, IQ)
Descriptive Statistics
The numerical, graphical, and tabular techniques used for organizing, analyzing, and presenting data
Type Function Examples
Graphs Provide a visual representation of the distribution Pie bar, histogram, polygon (Univariate)
of a variable or variables Clustered pie, clustered/stacked bar (bivariate,
nominal/ordinal scales)

More
Less
Related notes for HLTC15