Logic in Scientific Reasoning - PowerPoint PPT Presentation

1 / 168
About This Presentation
Title:

Logic in Scientific Reasoning

Description:

Compatibility with experimental laws/theories if two new hypotheses are ... 4. Test for reliability and validity. 1. Conceptualization ... – PowerPoint PPT presentation

Number of Views:924
Avg rating:3.0/5.0
Slides: 169
Provided by: roxanne5
Category:

less

Transcript and Presenter's Notes

Title: Logic in Scientific Reasoning


1
Logic in Scientific Reasoning
  • Definition of Science
  • Broadly speaking, science can be defined as a
    systematic study of nature and the rules which
    govern nature (and human behaviour or phenomena),
    to identify general statements of fact through
    which we can understand and interpret
    information.

2
Logic in Scientific Reasoning
  • Definition of Science
  • Science derives from Latin word scientia
    which means knowing.
  • Not talking specifically about particular subject
    areas, but an orientation/ way of looking at an
    area, a systematic approach in the pursuit of
    knowledge.
  • Basically any subject matter can be studied as
    science the essential point is the method used.

3
Scientific Method
  • The method distinguishes scientific from
    non-scientific enterprise
  • Here, one uses the tools of logic to assist in
    assessing whether a systematic and logical
    sequence has been followed in arriving at truth
    claims.
  • The scientific method can be described as the
    logic of science, ie, the principles that can be
    used to arrive at explanation of facts and so get
    knowledge of the world. It is a systematic
    approach that is used in the attempt to gain
    knowledge.

4
Scientific Method
  • The method combines both aspects of logic, ie,
    the inductive process and the deductive process.
  • The process is cyclical, with the method
    beginning and ending with what can be observed or
    data that can be gathered through means such as
    experiments, that is, through the use of
    induction.

5
Steps of the Scientific Method
  • Identifying a problem (remember that problem
    here means that which requires explanation, not
    the negative definition we generally have) cant
    have anything before a problem is identified.
  • Formulate a hypothesis cant be done without
    having collected some data that could be used to
    try to understand the events and so formulate a
    means of explanation

6
Steps of the Scientific Method
  • Collect additional data this generally involves
    preparing a research design to identify the
    sources of data and how that data will be
    collected. The problem at that stage is
    identifying what counts as relevant
    information/data. May be relevant for a
    hypothesis that you have not actually formulated,
    but not relevant to the one that you are
    presently using.

7
Steps of the Scientific Method
  • Test hypothesis how effectively does the
    hypothesis explain the facts that it was
    formulated to explain, as well as other facts?
    Additionally, how well can it predict future
    occurrences?
  • Draw conclusions - as to the efficiency and
    effectiveness of the theory in explaining and
    predicting events, and how the theory can be used
    to achieve some human goals (since that is often
    what is being aimed at, or results from
    scientific exploration).

8
Laws of Nature
  • The scientific method should yield a body of
    general statements of fact through which we can
    understand and interpret information.
  • These general statements of fact that we
    formulate are actually universal statements,
    because they should apply to all instances
    covered in the statement.

9
Laws of Nature
  • However, they are only provisional, in that there
    is always the possibility of finding evidence to
    disprove the statement.
  • These statements are also empirically based in
    that they depend on data gathered from
    observation or experiment.

10
Laws of Nature
  • These general statements are called laws of
    nature.
  • Laws of nature are derived from hypotheses that
    are formulated to understand the events
    occurrence.
  • Before we can have natural laws, we therefore
    need to have good hypotheses.

11
Evaluating Hypotheses
  • Relevance the hypothesis must not stray from
    the phenomenon which needs to be explained.
  • Testability Must always be able to test to see
    if the hypothesis is plausible by apply it to the
    data for which you need to be getting an
    understanding. Additionally, the test that is
    being done must be relevant to the hypothesis
    that is being tested.

12
Evaluating Hypotheses
  • Predictive and explanatory power distinction
    between the two is generally seen as temporal
    explanation referring to events that have already
    occurred prediction to those that have not yet.
    Yet at the same time, prediction presupposes that
    an explanation has already occurred.

13
Evaluating Hypotheses
  • Compatibility with experimental laws/theories
    if two new hypotheses are offered, want to accept
    the one that accords more with what has already
    been established. The idea here is that there is
    a constant accumulation of knowledge.
  • Simplicity may find that you have two theories
    that both function effectively when this
    happens, use the simpler of the two. Simpler the
    explanation, the less factors that youd be
    needing to take into account to give an account
    of the event.

14
Social Surveys
  • Research projects which use a questionnaire to
    collect standardized data from a large number of
    individuals.
  • Can be either Population or Sample surveys.
    Sample surveys are the most common
  • The collection of standardized data requires that
    the same questions be given to all respondents in
    the same order.

15
Types of Surveys
  • Factual Surveys Use to collect descriptive
    information. Example, Population census, The
    Survey of Living Conditions and The Labour Force
    Survey.
  • Attitude Surveys Carried out by opinion poll
    organizations, market researchers, etc.
  • Explanatory Surveys - Used to test hypotheses or
    to test and develop theories.
  • Common to all types, is the use of the
    Questionnaire as the instrument of data
    collection

16
Survey Design
17
The term Research Design can refer to
  • the planning of scientific inquiry
  • the design strategy for finding out something
  • the arrangement of the conditions of observations

18
All designs require
  • a precise determination of what you want to find
    out
  • the detailed specification of the most
    appropriate and effective way doing so

19
Factors determining Design
  • a. The purpose of the Study
  • b. The Time Dimension
  • d. Approach to the collection and handling of
    the data

20
The purpose of the Study
  • A particular research project can serve any
    one or a combination of the following purposes
  • Exploration
  • Description
  • Explanation

21
Classification According to Time
  • Time dimension speaks to the number of times
    participants will be observed in relation to a
    particular study
  • There are two approaches
  • - Cross-sectional researchers do a
    snapshot. One-time effort in gathering data.
  • - Longitudinal Permits the researcher to
    observe the phenomenon more than once

22
Correlational Design
  • The standard research design used in surveys is
    the correlational design
  • In this design constructs are measured
    independently of each other and then tested for
    associations.
  • Extraneous variables are controlled by including
    them in the study.

23
Causality
  • The Demonstration of Cause and Effect

24
Causation
  • Mill argued that every event that occurs has a
    cause this is called the Principle of Universal
    Causation.
  • But how do we identify causal relationships? If
    we see two events occurring together
    consistently, what right have we to assume that
    one is the cause of the other?

25
Explanation
  • Explanation implies delineating the causal links
    between and among variables.
  • To explain therefore involves asking two types of
    questions about any phenomena i.e. how? and/or
    why?
  • How questions are more easily answered than the
    why questions because of the nature of the types
    of answers required.
  • How questions can usually be answered using
    chronological accounts i.e. the sequence of
    events.

26
Explanation (contd)
  • Answering how questions allows us to decompose
    the structure of relationshipssimplify. e.g.
    traditional voting patterns
  • Why questions usually require introspection,
    rationalization and motivational answers.
  • Answers to why questions exist outside of both
    variables
  • Although one type of question is easer to answer
    than another both are necessary for adequate
    explanation.

27
Causality Assessment Criteria
  • There are four general criteria
  • 1. association
  • 2. time priority
  • 3. non-spuriousness
  • 4. rationale
  • interpretivists also rely on how they think
    things should be ordered.

28
Causality Assessment Criteria Association
  • For a causal relationship to exist there must be
    co-variation
  • Although association is necessary for causality
    the strength of the association does not increase
    or decrease causality, e.g. smoking and lung
    cancer
  • However if there is consistency i.e. if the
    association is present across different samples
    and contexts (established through replication)
    the greater the chance that there is a causal
    relationship.
  • this might be confounded by conditional
    relationships

29
Causality Assessment Criteria Time Priority
  • Cause must always precede effect.
  • This is not clear cut in all cases, it is
    difficult at times to determine which variable
    should precede the other, e.g. social class and
    educational attainment. How do we decide?

30
Causality Assessment Criteria Non-spuriousness
  • For two variables to be causally related the
    co-variation must not be the function of a third
    variable, i.e. a variable that is related to both
    variables, e.g. age and religiosity

31
Causality Assessment Criteria Rationale
  • Causality can only be established if theory is
    supported by empirical evidence. If there is any
    disconnect causality is questionable.

32
CausationMeaning of Cause
  • Necessary Condition can be stated in various
    ways
  • It has to be present for the event to occur.
  • If it is absent, the event will not take place.
  • If the phenomenon (effect) is present, the
    condition (cause) is present.
  • A is a necessary condition for B only if whenever
    B is present, A is present.

33
CausationMeaning of Cause
  • Necessary Condition
  • Will need several necessary conditions for a
    phenomenon to occur.
  • Example Plants will only grow if they are
    exposed to light.

34
CausationMeaning of Cause
  • Sufficient Condition can be stated in various
    ways
  • F is sufficient for H if whenever F is present, H
    is present.
  • If the condition is present, the phenomenon is
    present.
  • Once the condition is present, the event will
    occur.

35
CausationMeaning of Cause
  • Sufficient Condition
  • Example
  • The rain falling heavily is a sufficient
    condition for the road being wet, but can
    conceive of other ways in which the road could
    have become wet without the rain falling.

36
CausationMeaning of Cause
  • Remote vs. proximate cause
  • Remote cause looking at the more distant
    conditions that resulted in the events
    occurrence
  • Proximate cause the conditions that are in
    place just before the events occurrence
  • Remote and proximate causes will be linked by
    intermediate causes in what can be labelled a
    causal chain.

37
Causal Claims/Statements
  • These may be singular or general (we are usually
    more interested in the general)
  • Singular
  • The slippery road surface caused the accident
    today.
  • The tsunami in Asia on December 26, 2004 claimed
    500 000 lives.
  • General
  • Absence makes the heart grow fonder
  • Exercising reduces the risk of a heart attack.
  • Dirty campaigning wins elections.

38
Identification of Causal Relationships
  • Universality of causation if an event resulted
    from certain conditions being in place, we expect
    that this will happen again in the future, if the
    same conditions are in place.
  • This principle is the underlying basis through
    which we can develop general causal statements.

39
Identification of Causal Relationships
  • It is never possible to directly see causal
    relationships have to infer these relationships
    from observations.
  • Conclusions drawn from observations are
    potentially flawed as correlation does not lead
    to causation.
  • Correlation looking at the rate at which a
    property is seen among two populations/groups of
    events
  • The identification of a causal relationship
    generally starts with a correlation, but goes
    further by identifying a set relationship between
    the two, where the presence of one results in the
    presence of the other, or the absence of one
    results in the absence of the other.

40
Common Mistakes In Assigning Causal Relationships
  • May identify two events that are correlated, but
    may reverse the causal factors.
  • May identify a correlation between two events,
    but both events are the result of a common cause.
  • May have correlations that are purely accidental.
  • Note These errors will result in the Fallacy of
    False Cause.

41
Identification of Causal Relationships
  • Once a correlation has been identified, there
    needs to be a further exploration as to whether
    the events in question are in fact causally
    linked and, if they are, which is the cause and
    which is the effect.
  • This is usually the most difficult to achieve,
    because one needs to have strong support for the
    causal claim being made.

42
Identification of Causal Relationships
  • Several methods have been developed to ensure
    that when a causal statement is developed that it
    will actually hold true in most, if not all,
    cases.
  • John Stuart Mill developed five such methods,
    called Mills Methods of Experimental Inquiry,
    namely
  • Method of Agreement
  • Method of Difference
  • Joint Method of Agreement and Difference
  • Method of Residues
  • Method of Concomitant Variation

43
Mills Methods Method of Agreement
  • Looking for instances where two events
    consistently occur together, one being an
    antecedent to the other the antecedent condition
    being identified as the cause.
  • Example
  • In the case of chicken pox, you may notice that
    each time someone has chicken pox and someone
    touches her, the person also gets chicken pox.
    When you have seen this on several occasions, you
    can develop a general causal statement.

44
Mills Methods Method of Agreement
  • Each occurrence of this event with the conditions
    laid out would provide confirmation for the
    causal statement. In a case where you already
    have the cause, you are seeing a sufficient
    condition being identified.
  • If you already have the effect, trying then to
    identify what is the one condition that is
    present in all cases where the phenomenon occurs,
    so looking for the necessary condition.

45
Mills Methods Method of Agreement
  • Limitation
  • May not be able to narrow it down to one
    differentiating but common condition for all
    instances of the phenomenon.
  • Therefore may need to call on another method to
    supplement.

46
Mills Methods Method of Difference
  • You are looking at the conditions that are
    present when an event occurs and when it does not
    occur.
  • If there is one condition that is present when
    the event occurs and is absent when it does not
    occur, that condition can be seen as the cause of
    the event.
  • This is the method that is generally used in the
    controlled experiment. Trying to control every
    other variable want to see what will differ in
    the absence of one variable.

47
Mills Methods Method of Difference
  • Example
  • For the chicken pox, you may notice that you
    have three children playing with someone who has
    chicken pox. You watch carefully and no child
    touches the child that has chicken pox, and no
    child gets the virus, so in the absence of
    touching, the virus is not spread. So touching is
    the cause of the virus transmission.
  • Necessary condition is what is usually
    identified, or the basis for the necessary
    condition.

48
Mills Methods Method of Difference
  • Limitations
  • Cant control every other variable under normal
    conditions. Even a minor factor could change,
    which would make the use of this method
    problematic.
  • Very often, the cause of an event's occurrence is
    various factors being combined can then have
    more than 1 condition being identified as the
    cause, when all are actually required.

49
Mills Methods Joint Method of Agreement and
Difference
  • If you can find both instances when the condition
    and the phenomenon are present and when both are
    absent, there are stronger grounds for making a
    causal connection than using solely Agreement or
    Difference. Therefore, using both methods at the
    same time.

50
Mills Methods Joint Method of Agreement and
Difference
  • For the chicken pox, you may notice that you have
    three children playing with someone who has
    chicken pox. You watch carefully and two of the
    children touch the child that has chicken pox,
    and they get the virus. One child does not touch
    the infected child and does not get chicken pox.
    You can therefore conclude that chicken pox is
    transmitted through touching the infected person.

51
Mills MethodsMethod of Residues
  • There is complex mix of events occurring.
  • You already know the cause of some of the
    phenomena that are being exhibited. If you remove
    those causes and their phenomena, then you will
    have left some conditions and some effects.
  • A B C g h i
  • A has previously been identified as the cause of
    g
  • B has previously been identified as the cause of
    h
  • Therefore C is the cause of i

52
Mills MethodsMethod of Residues
  • Eat lunch, after which you start to feel ill. You
    have indigestion, vomiting and a rash. You had a
    tuna salad (which has mayonnaise), coconut water
    and a banana. You know from past experience that
    mayonnaise will give you a rash if it is combined
    with black pepper that coconut water can cause
    indigestion, then the banana is the probable
    cause of the vomiting.

53
Mills MethodsMethod of Residues
  • Limitations
  • May not be able to identify all the probable
    causes of the event in question.
  • You would have to go to another source to
    identify which of the possible causes that youve
    identified is the one that gave you the rash.
  • The causal link that is used to exclude the other
    occurrences from the equation have to be well
    established from other inductions how do you do
    this?

54
Mills MethodsMethod of Concomitant Variation
  • If you see a consistent variation between two
    phenomena, one increasing while the other
    decreases OR both increasing at the same time OR
    both decreasing at the same time, then have a
    basis for arguing that there is a causal
    relationship between the two events.
  • Example
  • Smoking causes lung cancer.

55
Mills MethodsMethod of Concomitant Variation
  • Limitations
  • Concomitant variation may be seen in situations
    where there is actually no causal connection
    between the two events.
  • Cant simply use two observations or instances,
    need to be looking at several cases.
  • When one mistakes correlation for cause, it is
    generally because of the misuse of this method.

56
Mills MethodsConcluding Remarks
  • Mill set out to try to show that we can arrive at
    certainty using induction by formulating his five
    methods. However, this cannot be achieved,
    because certainty, in the strictest sense, is not
    possible in induction.
  • He also searched in each of his methods to
    identify one cause, but many events have several
    factors which work together to lead to the
    occurrence of an event.

57
Mills MethodsConcluding Remarks
  • Mills Methods can be best seen as a means for
    testing hypotheses that we have formulated to
    explain particular phenomena we have observed or
    certain questions we want answered. This is
    because we will be limited in terms of the number
    of possible instances that we can identify to
    explain a particular occurrence youve already
    set limits once you start to use Mills Methods,
    they cannot be used in a vacuum.

58
MEASUREMENT
59
What is measurement?
  • The process of assigning numbers of labels to
    units of analysis in order to represent
    conceptual or variable categories
  • Scientific norms require that we fully describe
    our methods and procedures so that others can
    repeat our observations and judge the quality of
    our measurements

60
Steps in the measurement process
  • Conceptualization
  • 2. Operationalization
  • 3. Specification of levels of measurement
  • 4. Test for reliability and validity

61
1. Conceptualization
  • Creating factual or constitutive definitions of
    concepts
  • Define concepts in relation to other concepts.
    Ordinary language

62
2. Operationalization
  • Describing concepts in the language of
    measurement
  • Creating measurable definitions of concepts

63
3. Levels of Measurement
  • set of rules used in the labeling or quantifying
    of variables.
  • There are 4 levels, each assuming different
    interpretation of the numbers or labels assigned
    to the variables
  • Nominal Level
  • Ordinal Level
  • Interval Level
  • Ratio Level

64
Nominal variable
  • the term nominal means to name
  • measurement simply involves attaching names or
    labels to the variables
  • observations are merely classified into
    categories.
  • indicate whether things are the same or are
    different
  • numbers or labels are assigned to categories as
    codes for facilitate data collection and analysis
  • the only mathematical relationship that can be
    assumed is that of equivalence.

65
Nominal (cont.)
  • Cases placed in a given category must all be the
    same.
  • Also, the categories must be
  • exhaustive -sufficient categories so that all
    the items can fit into one of the categories
  • mutually exclusive -the items being classified
    must not fit in more than one category

66
Ordinal
  • Variables classified as ordinal also have the
    characteristics of mutually exclusive and
    exhaustive categories
  • In addition, this level of measurement has the
    additional feature of logical ordering of the
    attributes of the variables.
  • In other words, you can order or rank the
    variables under consideration

67
Interval level
  • This level has the qualities of nominal and
    ordinal measurements
  • In addition, there is equal distances (intervals)
    between the categories.
  • Example The difference between 20OC and 30OC is
    the same as the difference between 90O C and 100O
    C ( i.e. 10O C) degrees.
  • We can infer not only that 100OC is hotter than
    90OC degree
  • but also by how much because of a standard
    measurement.

68
RATIO LEVEL
  • This level includes all the features of the other
    levels of measurement
  • In addition, there is an absolute zero point.
  • Hence, it can be multiplied and divided.
    Example, income measured in dollars can be
    divided one into another to form a ratio.
  • Zero means absolutely nothing in this level of
    measurement

69
Reliability and Validity
70
Reliability
  • Reliability is concerned with issues of
    stability and consistency
  • It is the extent to which a measuring instrument
    produces the same result on repeated applications
    under similar conditions.
  • When repeated measures of the same thing gives
    identical or very similar results, the
    measurement instrument is said to be reliable.

71
Reliability Assessment Techniques
  • 1. The Test-Retest Method
  • 2. The Alternate-Form Method
  • 3. The Split-Half Method
  • 4. The Established Measures Method
  • 5. Inter-coder/Research Workers Readability

72
1. The Test-Retest Method
  • This is the simplest approach to assessing
    reliability
  • It involves testing/measuring the same persons or
    units on two separate occasions and then checking
    for statistical correlation between the two sets
    of scores.

73
Procedures
  • Test - Administer same test to some individuals
    on more than one occasion..
  • Compare - Scores of each individual on first
    testing are related to scores of second testing
    to provide a reliability coefficient.
  • Results - Coefficient can vary from 0 (zero),
    indicating no relationship between the sets of
    scores to 1 (one), indicating perfect
    relationship
  • Interpret - High coefficient close to 1 is
    desirable, since it is an indication of a strong
    relation between the scores or an indication that
    instrument is, indeed, measuring stable /enduring
    characteristics.

74
Advantages and Disadvantages
  • Advantages
  • Requires only one form of a test
  • Provides information as to test consistency over
    time.
  • Disadvantages
  • Affected by practice and memory
  • Influenced by events that might occur between
    testing sessions.
  • Requires the administration of two tests

75
2. The Alternate-Form Method
  • This approach reduces the likelihood that
    practice and memory will inflate reliability
    coefficient
  • Involves the use of two tests, the second being a
    parallel form of the first

76
Procedures
  • Test - Administer alternate forms of a test to
    same people
  • Compare- Compute relationship between each
    persons score on the two forms
  • Result - as above
  • Interpret- as above
  • It must be noted that this approach requires two
    forms of a test, which parallel one another in
    content and the mental operations required. In
    addition, items on one form must match items on
    the other form, with corresponding items
    measuring same quality or characteristic.

77
The Split-Half Method
  • a quick way of determining internal consistency
  • Procedures
  • 1. Split test in two halves (odd verses even
    or by random selection )
  • 2. Administer to two groups
  • 3. Relate scores of both groups
  • This approach determines whether each half of
    the test is measuring same quality or
    characteristic

78
The Established Measures Method
  • Use a standardized test/scale to measure the
    quality or characteristics of interest.
  • These are instruments for which reliability has
    already been established.

79
Inter-coder/Research Workers Reliability
  • Assesses the extent to which different
    interviewers, observers or coders, using the same
    instrument get equivalent results.
  • Involves the independent assessment and
    comparison of selected interviewers, coders or
    observers to determine consistency in judgments.

80
 Factors Contributing to unreliability of a
test
  • Familiarity with the particular test
  • Fatigue
  • Stress.
  • Physical conditions of the room in which the test
    is given
  • Health of the test taken.
  • Fluctuation of human memory
  • Amount of practice or experience by the test
    taken of the specific skill being measured.
  • Specific knowledge that has been gained outside
    of experience being evaluated by the test.
  •  A test that is overly sensitive to the above
    items is not reliable.

81
Validity
  • The extent to which a measuring instrument
    measures what it purports to measure
  • the truthfulness or accuracy of a measure.
  • Types
  • Criterion-related validity
  • Construct validity
  • Content validity

82
Criterion-related validity
  • Check instrument for its predictive power
  • Relate performance on the test to some actual
    behavior it is suppose indicate
  • Example- Drivers test. Relate performance on
    the written exam to use of the road use of turn
    signals, observation of signs, etc.
  • That is, relate test to performance criterion

83
Construct Validity
  • Established by relating a presumed measure of a
    construct or hypothetical quality to some
    behaviour or behaviour manifestation it is
    assumed to indicate
  • Relate performance on the test to some construct
    that the test score is assumed to indicate
  • Example- Self esteem scale A person who scores
    high on the test is assumed to have a high self
    esteem. Is that person extroversive, gregarious,
    highly motivated, etc.

84
Content validity
  • To what extent do the items on a test adequately
    represents all facets of the concept being
    measured?
  • Determined by ensuring that sample set (of items
    on test) is representative of actual set
  • A test in Social Research should cover all areas
    on course outline to be content valid

85
Dimensioning
86
Quality of life
  • Quality of Life There are many components to
    well-being. A large part is standard of living,
    that is, the amount of money and access to goods
    and services that a person has (easily measured).
    Additionally, the concept refers to freedom,
    happiness, and satisfaction with life are far
    harder to measure and could be more important.

87
Quality of life
  • The concept of quality of life incorporates two
    major dimensions
  • Objective Living Conditions This dimension
    concerns the ascertainable living circumstances
    of individuals, such as working conditions, state
    of health or standard of living.
  • Subjective Well-Being This dimension covers
    perceptions, evaluations and appreciation of life
    and living conditions by the individual citizens.
    Examples are measures of satisfaction or
    happiness.

88
Questionnaire Design
89
The Questionnaire
  • A questionnaire is a collection of questions and
    /or statements that is designed to collect
    information on a particular topic.
  • It is an instrument used by researchers to
    convert into data, information directly given by
    respondents.
  • In essence, it provides access to what is inside
    the person's head

90
  • The questionnaire facilitates the
  • measurement of what a person
  • knows - knowledge, information
  • likes dislikes - values, preference
  • thinks - attitudes, beliefs
  • experiences - past present
  • It is a useful alternative when direct
    observation is not possible.

91
  • This approach to data collection requires
  • that the respondent
  • co-operates in the completion of questionnaire
  • tells what is, instead of what he thinks ought to
    be, or what he imagines the researcher would
    like to hear.
  • knows how he feels or thinks in order to report.
  •  
  • It is possible therefore for the questionnaire to
    measure not
  • necessarily what a person likes, believes or
    thinks but what
  • he/she indicates in these regards.

92
  • The researcher must, therefore, pay attention to
    the following factors
  • He/she may not be able to provide answers to the
    questions posed - out of ignorance etc.
  • Respondent bias
  • Acquiescence the tendency to agree to statements
    despite the content of the statement
  • Social desirability respondents answer giving
    the socially or culturally correct answer rather
    than what they actually believe or feel.
  • Practice effects

93
Questionnaire construction
  • The structure of any questionnaire is determined
    by
  • Theoretical considerations
  • Method of data analysis
  • A questionnaire for a correlational or
    explanatory survey typically has the following
    types of items
  • Measures for the dependent variable(s)
  • Measures for the independent variable(s)
  • Background measures

94
Questionnaire construction (contd)
  • Question content
  • There are five types of question content
  • Behaviour what people do?
  • Belief what people think is true
  • Knowledge items that measure respondents
    knowledge of knowable facts or the accuracy of
    their beliefs
  • Attitude what people desire or find desirable
  • Attributes characteristics of people
  • Each type of question is specific to the
    characteristics that it is supposed to measure.

95
Questionnaire construction (contd)
  • Principles of item design
  • All items must achieve the following
  • Reliability
  • Validity
  • Adequate Discrimination
  • High response rates
  • Consistent interpretation across respondents
  • Relevance to the overall research endeavour

96
Questionnaire construction (contd)
  • Wording items
  • In wording items it is important to consider the
    following
  • Language simplicity
  • Length of the item
  • Avoiding double barrelled questions
  • Avoiding leading items
  • Avoiding negatively worded items
  • Avoiding ambiguous items
  • Avoiding prestige bias
  • Avoiding words (qualifiers) that will influence
    responses
  • The level of precision required to answer the
    item
  • The level of precision of the item and the
    knowledge required
  • Time and space requirements
  • The use of personal or impersonal wording

97
Types of Questions
  • Direct versus indirect (Specific vs. Non
    Specific)
  • a. Do you like your job? - direct (specific)
  • b. How do you feel about your job? -
    indirect (non-specific)
  • a. How you feel about teacher A? - direct
    (specific)
  • b.How do feel about class taught by teacher
    A? - indirect (non-specific)
  • Direct or specific questions may cause respondent
    to
  • become guarded or cautious and give less than
    honest
  • answers. Non-specific ones lead to desired
    information
  • with less alarm.
  •  

98
Types of Questions (contd).
  • Questions versus Statements - Can be a direct
    question as those types above (requiring a direct
    answer) or a statement requiring an optional
    response.
  •  Predetermined versus Response Keyed Questions -
    Answer all vs. answer those that are relevant.

99
  • 5. Do you drink alcoholic beverages?

    1. Never 2. Occasionally
    3. Frequently 4. Always
  • (If never, go to 6 and then terminate.
    Otherwise, skip to 7 and continue)
  •  
  • 6. Why dont you drink alcoholic beverages?
  • 1. Religious reasons 2. Health reasons 3.
    Others (Specify) ______

100
RESPONSEMODES
101
Structured Response (Close-ended)
  • Provide respondent with possible answers and ask
    him/her to choose the most appropriate option.
  • When the closed-ended format is used, the
    researcher should be guided by the following
  • - Response categories provided should be
    exhaustive
  • - Response options should be mutually
    exclusive
  • - There should be clear instruction to
    select the best answer
  • This format is respondent friendly and
    facilitates greater ease in the processing of
    data, since it can be transferred directly to
    computer. It however, limits the possible answers
    to those thought of by the researcher.

102
Structured Response (Close-ended)
  • For close ended the responses must satisfy the
    following requirements
  • Exhaustiveness
  • Exclusiveness
  • Balanced categories

103
Unstructured Response (Open-ended)
  • Researchers ask questions and allow respondents
    to provide answers
  • Exert control only in regard to the questions
    asked and the time and space provided.
  • Respondents give own answer, rather than just
    agreeing with those given.
  •  Format offers the respondent more flexibility

104
Disadvantages of Open-ended Format
  • Responses must be coded before processing - The
    coding process can be time consuming and can be
    quiet technical. It requires the researcher to
    accurately interpret the meaning of respondents
    give to responses. There is always the possibly
    of misunderstanding and researchers bias.
  • Respondents quite often provide answers that are
    irrelevant to researcher's intent.

105
Fill-in Responses.
  • This is transitional mode between structured and
    unstructured mode.
  • Respondents generate, rather than choose answers
  • Responses are, however, limited in range and
    length - often a single word or short phrase
  • Example What is your father's occupation?
  • The very wording of the question restricts the
    number of possible responses and the number of
    words.
  •  

106
  • Tabular Responses - Fill response into a table. A
    very convenient way of organizing complex
    responses.
  • Scaled Response - A structured response form.
    Respondents are asked to express endorsement or
    rejection of a given statement.
  • Numerical rating scales
  • These scales require respondents to give one
    response for each item
  • The resulting variable responses can be ordered
    from high to low
  • The numbers represent the intensity of the
    sentiment being expressed
  • Likert
  • Horizontal rating scale
  • Semantic differential scales
  • Vertical rating ladder

107
Fill-in responses (contd)
  • Scoring Out of 10
  • Ranking response Respondents are given some
    statements, etc. and asked to rank according to
    some criteria.
  • Checklist Response (Multiple response format) -
    Respondents choose all possible answers from a
    number of options given to him

108
Fill-in responses (contd)
  • Binary choice formats
  • Dichotomous
  • Paired comparisons given these two choices which
    do you consider to be more important?
  • Multiple choice formats
  • The respondent is asked to select one response
    from a set of responses.
  • Multiple nominal categories
  • Multiple ordinal categories
  • Multiple ordered attitude statements

109
Fill-in responses (contd)
  • Non-committal responses
  • No opinion
  • Dont know
  • Middle non-committal answer

110
Fill-in responses (contd)
  • The number of response categories
  • The fundamental concern is that there are
    response categories that the respondent can
    comfortably represent themselves. Secondly this
    depends on the ability of the respondent to
    answer the item and the extent to which they can
    do so.
  • Dichotomous
  • Five point scales information about intensity,
    extremity and direction
  • Longer scales greater discrimination and
    therefore finer details

111
Development Issues
  • In constructing the Questionnaire, the
    researcher
  • should always consider the following factors
  • Format Wording
  • Precision Questions should be clear and
    unambiguous
  • Concision Items should be as short as possible
  • Relevance Question should all be relevant and
    necessary
  • Double-barreled Questions Each question should
    should attempt to measure only one variable at a
    time
  • Biased Items/Terms Should not use leading
    questions
  • Negative Items Questions should be in positive
    form
  • Abbreviations and Jargons These should always
    be avoided

112
Development Issues (contd).
  • Format Layout
  • Uncluttered Items should be well-spaced/
    spread-out
  • Order Items should flow in a logical order. The
    ordering of questions affects the quality of
    responses
  • Length Should not be too many items
    Instrument shouldnt be too long
  • Personal Information Request only when required
  • Instructions Always provide adequate
    instructions both general and specific.

113
Pilot or Pre-testing
  • Questionnaire development

114
Why Pre-test?
  • It is human to err irrespective of how
    systematic or careful we are in the questionnaire
    design process we will more than like make
    errors.
  • The standardized questionnaire is inflexible.
    Once the instrument is developed and data
    collection has started any mistakes/errors cannot
    be corrected. If modifications to the
    questionnaire are made during data collection,
    any data collected before the changes will become
    useless.
  • To determine the length of time the questionnaire
    takes to be completed.

115
Conducting the Pre-test
  • Use respondents that are part of the population
    but not part of the sample. The basic requirement
    is that the questionnaire should be relevant to
    those responding.
  • The interview should be conducted in conditions
    as similar to those that the actually interviews
    will be conducted under. That is, similar
    settings and the interviewers that will
    eventually conduct the interviews should be used.
  • Test in waves that is, after each set of
    revisions to the instrument, the revised
    questionnaire should be re-tested on a new set of
    respondents. These waves of testing should be
    continued until the instrument is as clean as is
    possible.

116
Sampling and
  • Sampling distributions

117
What is Sampling?
  • A process for identifying and selecting elements
    for observation (Babbie, 2000)
  • The selection of units of observation in such a
    manner that a researcher can make relatively few
    observations yet being able to generalize to the
    larger population
  • A systematic way of deciding what or whom to
    observe when limited resources dictate that the
    few instead of the many be observed

118
Why Sample?
  • Factors justifying the observation of a sample
    rather than the entire population
  • Cost constraint
  • Time constraint
  • Timely Results
  • Accuracy Planning and logistics more manageable
  • Human Resource constraint
  • Practicability e.g. Population may be
    inaccessible

119
Some Important Concepts/Terms
  • Representativeness typical of the population. A
    sample is representative of the population if the
    aggregate characteristics of the sample closely
    approximate those same aggregate characteristics
    of the population
  • Equi-probability the probability or chance for
    being selected is the same for all members of the
    population. Sampling methods which ensures
    equi-probability are classified as Random
    Sampling Techniques

120
Some Important Concepts/Terms
  • Bias giving a quality or characteristic
    more/less attention or emphasis than it merits.
  • A biased sample is not typical/representative of
    the population
  • Factors contributing to bias in a sample
  • - Convenience tendency to include units that
    are easily accessed
  • - Personal biases
  • - Ignorance about composition or description
    of population
  • - Flawed Method

121
Some Important Concepts/Terms
  • Population Totality of items or units about
    which the researcher wants information
  • Sample Frame An accurate specification of all
    units of interest to a particular study. It is an
    operational definition of the population under
    consideration. A list of all units of interest.
  • Sample A subset of the population drawn from
    the sample frame. A representative smaller group
    that is systematically selected from the
    population
  • Element A unit of sample drawn from the sample
    frame of a particular population. Member of
    sample

122
Some Important Concepts/Terms
  • Sampling error this is the difference between
    the value of a statistic and the value of the
    corresponding population parameter.
  • Periodicity Where there are periodic/cyclical
    patterns in the population that correspond to the
    sampling intervals (a problem specific to
    systematic random sampling which results in a
    biased sample).
  • This is caused by the arrangement of the items in
    the population is defined by some characteristic.

123
Two types of Sampling
  • Probability
  • - Elements are drawn based on chance/random
    procedures
  • - Every member of the population has known
    (non-zero) probability of being selected
  • Non-probability
  • - Elements are not chosen by chance/randomly
  • - Determined on the basis of expertise,
    personal judgment, knowledge or convenience

124
Non-probability Sampling
  • In many research situation, the enumeration
  • of the population elements (basic requirement
  • of probability sampling) is difficult or
  • impossible
  • Other times, a representative sample is not
    appropriate given the aim/purpose of the research
  • In these instances, a non-probability technique
    will suffice

125
Probability Sampling
  • There are four main types of probability samples.
    The choice between these depends on the nature
    of
  • The research problem
  • The availability of good sampling frames
  • Money
  • The desired level of accuracy in the sample
  • The method by which the data is to be collected
  • Four Types
  • Simple random sampling
  • Stratified random sampling
  • Systematic sampling
  • Multistage Area/Cluster sampling

126
Simple Random Sampling
  • All members of the population have an equal and
    independent chance of being included
  • Define population
  • List accessible members of population (complete
    sample frame)
  • Decide on the required sample size
  • Select sample by employing chance procedure (e.g.
    table of random numbers)

127
SRS Table of Random Numbers
  • Table containing columns of digits that have been
    computer generated.
  • Assign each member of population a distinct
    identification number
  • Use table of select systematically, subjects to
    be included in sample
  • Customary to determine by chance the point at
    which table is entered

128
SRS
  • Disadvantages
  • Requires an unbiased sample frame
  • Impractical when surveying populations in diverse
    geographical areas

129
Systematic Sampling
  • Involves drawing a sample by taking every
  • Kth case from a list of the population
  • First decide on a number of subjects in the
  • sample (n)
  • Since the total number in the population (N) is
  • known, divide N by n and determine the
  • sample interval (k) to apply to the list

130
Systematic Sampling
  • First number is randomly selected from the
  • first k member of the list (N.B. if this is
    not done the list will be exhausted before the
    sample is selected) and every Kth member of the
    population is selected for the sample
  • If a fraction is obtained truncate the number and
    proceed. If the number is rounded instead the
    list may be exhausted before the sample is drawn.
  • Pop 500 desired sample 50
  • kN/n
  • 500/50 10

131
Systematic Sampling
  • Start near the top of the list and select the
  • first case randomly from the first 10 cases and
    then every 10th case thereafter
  • Differs from simple random because
  • choices are not independent. Once first
  • case is chosen, all subsequent cases are
  • automatically determined.

132
Systematic Sampling
  • If the original population lists is in random
  • order, then systematic sampling would yield a
    sample that could be reasonably substituted for a
    random sample
  • If list is alphabetical or otherwise structured,
    (e.g. people are positioned in population
    according to a given characteristic) the sample
    may be biased (Periodicity).

133
Systematic Sampling
  • Advantage
  • Simplest method to select a random sample
  • Disadvantage
  • Requires an unbiased sample frame
  • Periodicity

134
Exercise
  • 1, 2,3, 4, 5, 6, 7, 8, 9, 10,11,12,13,14,15,16,17,
    18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34
    ,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,5
    1,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,
    68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84
    ,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,
    101,102,103,104,105,106,107,108,109,110,111,112,11
    3,114,115,117,118,119,120,121,122,123,124,125,126,
    127,128,129,130,131,132,133,134,135,136,137,138,13
    9,140,141,142,143,144,145,146,147,148,149,150.

135
Stratified Sampling
  • Use when population contains a number of
    subgroups/strata that may differ in the
    characteristics being studied
  • First identify the strata of interest and then
    draw a specified number of subjects from each
    strata
  • This approach improves representativeness and
    facilitates the studying of differences between
    subgroups

136
Stratified Sampling
  • To Stratify a sample
  • Select the stratifying variable
  • Divide the sampling frame into separate lists-one
    for each category of the stratifying variable
  • Draw a systematic or SRS of each list

137
Stratified Sampling
  • Advantages
  • It produces more representative samples
  • Disadvantage
  • More complicated than SRS and systematic random
    sampling
  • Sample frame must contain information on the
    stratifying variable
  • Requires an unbiased sample frame

138
Multistage Cluster Sampling
  • Using this method a (final) sample is obtained by
    drawing several different (intermediate) samples.
  • Procedure
  • Divide population into broad groups clusters
  • Select a SRS of these groups
  • Within these selected groups obtain sample frames
    of each
  • Select a SRS within these groups, etc.

139
Multistage Cluster Sampling
  • Advantages
  • Does not require an unbiased sample frame
  • Can be used to sample geographically diverse
    populations
  • Disadvantages
  • Complex
  • Expensive

140
Steps in the Sampling Process
  • Define the population Identifying all the
    elements about which the research wants
    information
  • Determine Sample Frame -Identify the accessible
    population
  • Select Sample Systematically Select a
    representative smaller group from sample frame.
  • Observe - Make observation of this smaller group
  • Generalize - Generalizing the findings to the
    larger population

141
Sample Size
  • Sample size is determined by two factors
  • The degree of accuracy that is required
  • The level of variation in the population across
    the variables being studied

142
Sampling distributions
  • Up until this point the discussion about
    statistics has been about those from a sample. In
    our estimation of the various means and standard
    deviations it was assumed that these sample
    statistics were similar to those in the
    population. However this assumption is not
    necessarily true. To begin our discussion about
    how this is the case we need to understand some
    basic statistical terms and their relevance to
    sampling distributions.

143
Sampling distributions
  • A population is any set of measurements that can
    be made of a random variable real or
    hypothetical.
  • A sample is a subset of these (actual)
    measurements.
  • In an effort to summarize the characteristics of
    samples various numerical descriptive measures
    called sample statistics were calculated from the
    sample measurements and used as afore mentioned.
  • Since sample statistics are singular numerical
    values they are also referred to as point
    estimates of population parameters.

144
Sampling distributions
  • Populations characteristics are also summarized
    by similar numerical descriptive measures called
    population parameters however these are usually
    unknown because it is too expensive to measure
    every possible value of a random variable or it
    is impractical to do so for some other reason.
  • Consequently sample statistics are used as
    estimates of population parameters because it is
    cheaper to make measurements of a few of the
    values that a random variable can take.

145
Sampling distributions
  • Specifically sample statistics are used to make
    inferences about population parameters either in
    the form of (1) estimates of the actual
    population values or (2) to formulate a decision
    about the value of a population parameter.
  • However this is problematic because the sample
    statistics vary from sample to sample depending
    on the values of the random variable that are
    sampled.

146
Sampling distributions
  • The solution is to create a probability
    distribution (sampling distribution) of all the
    values of the statistic produced from the various
    samples that can be drawn from the population of
    measurements of the random variable.
  • This probability distribution could then be used
    to evaluate the reliability of the inferences
    made about the population parameters. Of course
    we are assuming that each sample consists of
    values typically found for the random variable or
    in other words each sample is representative of
    population of values it is drawn from. The most
    common mechanism used to ensure
    representativeness is to select the values of
    each sample by a random method.

147
Sampling distributions
  • Understanding sampling distributions is therefore
    important because it facilitates the
    comprehension of the process of statistical
    inference in particular an intuitive appreciation
    of the reliability of the inferences made using
    sample statistics.

148
Sampling distributions
  • Properties of Sampling Distributions
  • The mean of a sampling distribution is normally
    distributed.
  • The mean of the mean sampling distribution is
    equal to the population mean

149
Sampling distributions
  • If the mean of the sampling distribution is not
    equal to the mean of the population then the
    sample mean is said to be biased.
  • The standard deviation (or standard error) of the
    mean sampling distribution is

150
Sampling distributions
  • The mean distribution can be converted to the
    standard normal distribution by using,

151
Sampling distributions
  • where µ the mean of the random variable
  • n the sample size
  • s the standard deviation of the random variable

152
Sampling distributions
  • From the properties listed above two theorems
    have been derived, the first says that if a
    random sample of n observations is selected from
    a population of measurements from a random
    variable that is normally distributed the mean
    sampling distribution that will be produced from
    this population will also be normally distributed.

153
Sampling distributions
154
Sampling
  • Sample size
  • Before sample size selection, the following must
    be considered
  • Level of confidence this is the risk of error
    the researcher is willing to accept in the study.
    This in turn depends on
  • Time
  • Money/resources
  • Consequences associated with drawing incorrect
    conclusions

155
Determining Sample Size
  • The most common levels of confidence used are 95
    and 99.
  • Confidence interval this is the level of
    sampling accuracy the researcher will have.
  • The type of variable or variables being studied

156
Sample Size
  • Categorical variables require different sample
    sizes than do metric level variables. Generally
    categorical variables require smaller sample
    sizes than metric variables.
  • Additionally it is easier to estimate sample
    sizes for categorical variables. That is, in
    order to calculate the sample size for metric
    variables, the standard deviation of a previous
    sample is required.
  • n z2pq/e2 , where e z v(pq/n)

157
Sample Size
  • This standard deviation is usually obtained from
    one of two sources either from previous research
    literature using the same population or from the
    sample used in the pre-test of the instrument.
  • n (zs/e)2
  • Where e z (s/vn)

158
Sample Size
  • The size of the population
  • If the population has less than 100,000 elements,
    it is considered to be a finite population (or if
    n /N 0.05) and the finite population correction
    factor must be applied to the calculation of the
    standard error. The resulting samples are smaller
    than those selected from larger populations.
  • Finite population correction factor v (N n)
    / ( N 1)

159
Exercise
  • An alumni association wants to estimate the mean
    debt of this years college graduates. It is known
    that the population standard deviation of the
    debts of this years college graduates is 11,800.
    How large a sample should be selected so that the
    estimate with a 99 confidence level is within
    800 of the population mean?
  • A political party wants to estimate the
    proportion of voters who disapprove of a
    candidate they have put to run in a particular
    constituency. The party wants the estimate to be
Write a Comment
User Comments (0)
About PowerShow.com