SWP32RES RESEARCH FOR SOCIAL WORK PRACTICE B

1
SWP32RES RESEARCH FOR SOCIAL WORK PRACTICE B

• LECTURE THREE - Overview of the Research Process

2
The Social Work Research Process

• 1) Identify the Research Problem Area
• 2) Identify Personal Motivation for Interest
  in this Problem Area
• 3) Formulate Focussed Research Question/s or
  Hypotheses

3

• 4) Review the Literature
• 5) Reformulate Research Question/s or
  Hypotheses
• 6) Develop a Plan for the Research Study

4
1. Operationally define key terms,
concepts and variables.2. Decide on research
approach - quantitative, qualitative or a
combined approach.3. Decide on research
design. 4. Define and decide upon access to
sample and sample size.
5

• 5. Decide on data collection method and
  instruments (questionnaires, interviews,
  secondary data analysis or observation)
• 6. How the data is to be analysed.
• 7. Staging and timing of the study.
• 8. Costing of the study.

6

• 9. Pretesting and Piloting of the data
  collection instrument/s.
• 10. Write up research proposal.
• 11. Obtain necessary ethics clearance.

7

• 7) Collect the Data
• 8) Analyse the Data (includes the data
  reduction (whether quantitative or qualitative)
  and statistical analysis)
• 9) Write up the Research Findings
• 10) Disseminate the Findings
• 11) Implement and Utilise the Findings

8
Abuses of Research Ethics

• 1) Experiments on Jewish inmates of Nazi
  concentration camps during World War 11
• 2) The Tuskeegee experiment
• 3) Milgram studies
• 4) The Laud Humphries study

9
Responses to ethical abuses

• 1) World Medical Associations Helsinki
  Declaration (1964) 2) This Declaration has
  been directly translated into the Australian
  National Health and Medical Research
  Councils (NHMRC) guidelines on human
  experimentation

10

• 3) A.A.S.W. (The Australian Association of
  Social Workers) which has a section on
  research ethics on the latest version of its
  Code of Ethics (1999)

11
COMMONLY ACCEPTED PRINCIPLES OF ETHICAL RESEARCH

• 1) VOLUNTARY PARTICIPATION INFORMED CONSENT
• 2) PROTECTION FROM HARM
• 3) ANONYMITY CONFIDENTIALITY

12

• 4) PREVENTION OF DECEPTION
• 5) PROPER CREDIT FOR RESEARCH ENDEAVOURS

13
Bradshaws Taxonomy of Social Need

• In practice, four definitions of Need
• 1. Normative Need (defined by experts/consensus)
• 2. Felt Need (people say they WANT it)
• 3. Expressed Need (WANT turned into DEMAND)
• 4. Comparative Need (Individual or service
  provision differences across similar areas)
• Interrelations Degree of Support (O/Head) -
  apply all four to your project

14
Computer Lab Info

• Using SPSS (Statistical Package for the Social
  Sciences version11.5)
• Allows simple thru to complex analysis of your
  data Tables Graphs (Visual can enhance
  decision-making understanding)
• In social sciences we are more likely to use
  categorical data data (plural) is discrete (eg
  sex, race, religion, marital status, employment
  status, etc) rather than continuous, (eg, age,
  years of education, or score on a test).
• Graphing these discrete use a Bar Chart or Pie
  Chart, cf continuous data using histogram, line
  graph (polygraph) or scatterplot.

15
Exploratory Data Analysis (EDA)

• EDA appropriate for both qualitative and
  quantitative data. Answer the question What is
  the data telling us about ..?
• Searching for ways of revealing meaning in the
  data. The more we know about the data the more
  effectively we can use it to refine practice and
  theory.
• EDA applied to data that is Univariate,
  Bivariate, Multivariate, and Causal Analysis
  (Causal Pathways very in these days eg child
  development field, antenatal pathways etc..)
• See O/Head arrow diagrams simple but effective
  conceptual framework tool.

16
EDA contd

• Data made up of variables A Variable is a set of
  values each of which represents an observed value
  for the same characteristic for one of the cases
  being used in the research, eg income, age, sex,
  and so on.
• When all observed values put together and ranked
  we have a distribution from lowest to highest.
  That is, cases are distributed across a range of
  values.
• We want to visualise the shape and spread of that
  distribution eg bell curve graphs
• Note assume we have ungrouped data (unit record
  data). If using secondary data source, data may
  already be grouped.

17
Measures of Central TendencyDESCRIPTIVE
STATISTICS

• Shape of distribution symmetric or skewed? The
  symmetric bell-shaped normal distribution
• Spread of values? Are there unusually high and
  low values or all clustered?
• Mean (Interval or ratio data is the average),
  median (ordinal data half above and below, or
  50th ile), mode (nominal data most frequently
  occurring score or value), SD (std deviation)
• If mean, median and mode all the same then the
  data are symmetrical.

18
Data Levels of Measurement

• 1. Nominal data (also referred to as Categorical,
  Discrete, Dichotomous) has mutually exclusive
  categories. We often look at relationship between
  categorical variables by Cross-tabs and bar
  charts (in SPSS). Usually only a small range of
  values. Can assign a numerical code to count
  categorical data.
• 2.Ordinal data similar to nominal data but has
  some order or ranking to it, eg client
  satisfaction, level of education, agreement
  ratings (SD thru SA Likert scale)
• 2.1 Nominal and Ordinal data sometimes referred
  to as Qualitative data, whereas Interval and
  Ratio referred to as quantitative data.
• 3. Interval Data (also known as continuous,
  metric). Values are different in magnitude but
  the difference is not meaningful, eg Celsius or
  IQ scale. Someone with score of 100 not twice as
  bright as score of fifty.
• 4 Ratio Data Like Interval data (continuous
  )but difference in values is meaningful eg age.
  Age 20 cf age 10 ratio is 21. Scatterplots
  often used to visually illustrate relationship
  between two continuous variables.

19
Descriptive data analysis

• Overall in social work research, we are
  interested in individual variables and their
  shape and size as well as the relationships
  between variables, how they co-relate or co-vary
  (as one increases so does the other which
  variable affects the other), whether they form
  groups, and if so can we predict group
  membership, and so on. This involves the use of
  bivariate and multivariate data techniques.
• Descriptive data is very useful in describing
  characteristics of our sample, but limited
  because univariate. We need bivariate data to
  get more specific meaning as to whether there is
  a relationship between variables, and if so, how
  strong that relationship is, and its direction.
  Eg, the impact of different interventions.
  Descriptives only tell us so much. We then use
  bivariate data to get more detailed understanding
  of impact of intervention.
• Must also be very clear on what is the question
  we are trying to answer. Descriptives help us
  summarise the raw data and make it comprehensible
  in terms of the research question.
• Also, important that we use the appropriate
  measure for the type of data we have. That is,
  discrete or continuous data. Note Alston
  Bowles use just two terms 1. DISCRETE covers
  all nominal and ordinal data, and 2. CONTINUOUS
  covers all Interval and ratio data.
• We should stick to these two terms as well -
  Alston Bowles Ch 14, p234 on. V/good for
  assignment.

20
Normal Distribution Descriptive Stats
(continuous /interval data only)

• Symmetrical, bell-shaped data referred to as
  PARAMETRIC DATA (cf non-parametric eg
  Correlation, see later slide)
• (see graph O/Heads normal, skewness, kurtosis,
  and Standard Deviation)
• Assumptions of parametric data
• 1.normally distributed data
• 2. Homogeneity of variance
• 3. interval level data
• 4. Independence
• Do a FREQUENCY DISTRIBUTION ( diagram) to see if
  our data violates assumps, or is from non-normal
  distribution, re appropriateness of descriptive
  stats (Mean, median, Mode, percentages, and
  graphs). For example, average income data often
  distorted by very high and very low income
  earners.

21
Descriptive stats

• Plot a histogram to look at the distribution of
  data, and can overlay normal curve.
• Check for extreme values or outliers (eg Kerry
  Packers very high income) Can distort the
  Mean. Remove or transform data.
• Discrete data (Categorical / nominal /
  Non-interval data) not approp for mean, or
  Median. Mode ok for nominal data, and Proportions
  , and Bar or Pie charts. Also Frequency tables
  appropriate where nominal data has small number
  of categories eg sex, marital status, etc..
• Some ordinal data (eg Likert scales) can be
  treated as interval to obtain mean, eg QA.
• Need to SELECT each Intervention group and run
  Descriptives on each, for comparison. And on
  variables within each intervention group, eg
  gender. Or, cross-tab variables of interest and
  use control variable (layer) such as intervention
  group or gender and so on.

22
Non-parametric data and tests

• Less, or more relaxed, assumptions about interval
  data data not need to be normally distributed
• Tests based on ranking of data rather than actual
  data itself.
• Eg, Correlation Pearson parametric or
  Spearmans non-parametric)