Quantitative approaches

1
Quantitative approaches

• J.Maclean
• SHCS

2
Quantitative research

• Deals with numeric measurement (ie. quantities)
• Aims to test hypotheses, identify numerical
  difference between groups
• Often on a larger scale than qualitative research
• Often aims to infer ie to project findings from
  a sample on to a population

3
Quantitative research

• a formal, objective, systematic process in which
  numerical data are used to obtain information
  about the world. This research method is used to
  describe variables, examine relationships among
  variables and determine cause-and-effect
  relationships among variables.
• Burns Grove (1995)

4
Design

• The research design provides the plan for
  answering research problems.
• The design becomes the vehicle for hypothesis
  testing or answering research questions. The
  design involves a plan, structure and strategy.
• LoBiondo-Wood Harber (1998)

5
Design

• Design may be broadly divided into
• Experimental designs
• true experimental
• quasiexperimental
• Nonexperimental designs
• Survey studies, observational studies
• Studies of relationship / association

6
Experimental design

• An intervention is manipulated by the
  investigator, under defined and controlled
  conditions.
• The interventions effect is then assessed
  (effect of the independent on the dependent
  variable)
• Comparison is made between the group exposed to
  the intervention, and a control group which has
  not been exposed.

7
Experimental design

• Randomisation of sample is required that is
  assignment of participants to experimental or
  control group using a randomising procedure.

8
Experimental design

• Involves use of pre- and post-test measurement.
• Cause and effect relationship may be tested
• Independent variable ..
• has a presumed effect on the dependent variable
• manipulated by the researcher
• demonstrates cause

9
Experimental design

• Dependent variable
• the consequence or presumed effect that varies
  with the change in the independent variable
• is not manipulated
• is observed/measured and assumed to vary with
  changes in the independent variable
• demonstrates effect

10
Experimental design

• May be known in health research as randomised
  controlled trial
• RCT quantitative, comparative, controlled
  experiment, where investigators seek to measure
  and compare outcomes of 2 or more clinical
  interventions.

11
RCTs

• Participants allocated at random to receive one
  of interventions
• Control may be standard practice, a placebo, no
  intervention.
• Eg. Study of new NSAI drug
• Patients with arthritis randomly allocated to
  conventional drug, or new drug
• Effect quantified.

12
Quasi-experimental design

• Used when full experimental control is not
  possible
• Either control is not possible or randomisation
  is lacking
• Does involve the use of an experimental group
• Used to determine cause and effect, but
  confidence in the results is weakened

13
Non-experimental design

• Survey
• Alternative approach if experiment inappropriate
• Types of variables of interest can be opinions,
  attitudes, facts
• Data collection from field of interest by methods
  such as questionnaire, interview.

14
Survey

• Prospective/retrospective/longitudinal
• an economical method to obtain a large amount of
  data
• If sample is representative, data can give a
  reasonably accurate picture of the population
• Information obtained may be superficial
• Large scale and longitudinal studies may be
  logistically hard, and expensive

15
Design

• Non-experimental association/correlation
• Correlational design is used to examine the
  relationship between two or more variables
• No attempts made to determine causation
• Researchers investigate the strength of
  relationship between the variables

16
Prediction

• May want to see if any variable has power of
  prediction
• Following correlation study, select variables for
  a regression study
• An independent variable is regressed on a
  dependent variable
• Indicates whether one can predict another

17
Samples

• The sample relates to the population, which may
  be too large to study in its entirety.
• Population aggregate of people/objects which
  are the focus of interest, for example

18
Sampling

• In practice we cannot study the entire population
  so get our information from a selection of units
  from the population
• Sampling is the process of identifying a suitable
  sample in order to determine characteristics of
  whole population
• Statistical inference is the process by which
  informed estimates of the populations
  characteristics are made.

19
Sampling

• Aim is to draw a representative sample so we can
  make generalisations, inferences about the target
  population
• A miniature version of the overall population
• An unrepresentative sample reduces the validity
  of the study

20
Sampling techniques

• Random sampling each unit of the population has
  a calculable chance of being selected
• Non-random sampling units of the population are
  selected on the basis of some kind of judgment

21
Random sampling

• Unrestricted numbers/ allocated to entire
  population random numbers generated to select.
• Each unit returned to population before each draw
  so a number can be selected several times
• Equal chance of selection

22
Random sampling

• Simple numbering as in unrestricted, but
  selected units are not returned to population
  before next draw
• Equal chance of selection

23
Systematic and stratified random sampling

• Systematic choose start point at random then
  select every nth unit
• NB is list truly random eg not in alphabetical,
  chronological order?
• Stratified population subdivided into strata
  eg malefemale
• Select randomly from each but in proportions
  found in population

24
Cluster sampling

• Deals with dispersed population
• Initially area to be covered is grouped together
  into clusters.
• Clusters are then randomly sampled

25
Non-random sampling

• Convenience sampling selection of units of
  population which are accessible/close at
  hand/available at a given time period.
• Quota sampling like stratified sampling but
  with no random element
• Purposive sampling selects sample which fulfils
  certain criteria but is not randomised

26
Sample size

• Optimum sample size is that which allows correct
  inferences to be drawn with regard to population.
• Larger the sample, closer to measuring population
• BUT may be impractical, costly, unethical
• Statistical techniques will give indications of
  accuracy

27
Sampling error

• This is probability that any one sample from a
  target population, is not fully representative of
  that population.
• It shows the potential mismatch between
  characteristics of sample versus the population -
  an inherent uncertainty in any sample

28
Statistical power

• Measures likelihood of a study to produce a
  statistically significant difference between
  groups/subjects
• If power is low then results may not give a
  true picture
• Power calculation helps avoid a Type II error

29
Analysing quantitative data

• Numeric data
• Statistical approaches
• Descriptive and often inferential statistics
• Parametric or non-parametric approach?