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