RM (Research Methodology)
:: Survey Research :: Medium: Personal (response rate = 30-35%)
Telephone (25-30%) Computer (0.5%) Mail/post (2%) :: To decide
which type to use: Do the respondents & interviewers have
common understanding of the information sought? Is the respondent
able to provide a truthful answer? Is the respondent willing to
provide a truthful answer?
RM (Research Methodology)
:: Survey Research :: Criteria for selecting method: Complexity
of questions Amount of data/time Data accuracy required sensitive
questions, interviewer effects, error sources Cost Sample control
Response rate :: Unstructured indirect interview: Projective
techniques, underlying beliefsEg: Sentence completion,
picture/story completion
RM (Research Methodology)
EXPERIMENTAL RESEARCH :: Experimentation: Experimentation is the
manipulation of one or more variables to measure their effect on
one or more variables. That is, cause and effect studies :: Types
of variables: Independent Dependant variable :: Treatment group:
Ample exposed to manipulation of independent variable
RM (Research Methodology)
EXPERIMENTAL RESEARCH :: Control group: The independent variable
is unchanged. To ensure that the change in independent variable is
caused by independent variable, other variables need to be
controlled/measured in terms of their effects. Randomization
Matching :: Experimental design features are: Error control
Time/Sample/Data Monetary requirements
RM (Research Methodology)
EXPERIMENTAL RESEARCH :: Symbols:: MB MA X R = = = =
Pre-measurement Post-measurement Treatment Random selection
RM (Research Methodology)
EXPERIMENTAL RESEARCH :: Basic designs :i) After only design : :
: R1 R2 MB MB1 MB2 MB X X MB1 MB2 X X X MA MA1 MA2 MA1 MA2 MA3 MA4
X X X X MA MA MA1 MA2 ii) Before After Design iii) Before After
with design control
(iv) Simulated Before After design : R1 R2 (v) After only with
control : R1 R2 (vi) Solomon 4 group control :R R R R
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN :: Statistical design: They permit
measurement of effect of more than 1 independent variable. It
allows researchers to control for specific extraneous variables
that may affect results. It is more economical :: Types :
Randomized Blocks Design Latin Square Design Factorial Design
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN (i) Randomized Blocks Design (RBD)
Completely random design assumes that each treatment group is
similar in reaction to independent variable and is similar
initially in terms of dependent variable. This may be invalid.
Hence RBD is used. In RBD, experimental units are blocked/grouped
on the basis of blocking / grouping variable. Hence,
experimental/treatment group & control group are better
matched.1 IV 1 IV 1 DV
1 EV 1 EV
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN Q1. Hoppers Hop has 800 outlets all
over India. They are planning a major sales promotion during the
peak Diwali buying period. To determine the costing and maximized
profitability, they have identified three different price levels
P1, P2, P3 at which they can offer their goods. The sales volumes
at their outlets vary from Rs.3,00,000 pm to Rs.30,00,000 pm. They
are apprehensive about trying out all the 3 prices at each location
as it may disturb their customer base. Hence, they have approached
you to design an ext which would minimize customer disturbance, and
at the same time provide solution profit maximization during
Diwali. Additional information indicates that sales of the product
specified are a function of the total sales at that particular
store. You have identified 27 stores for your experiment and the
process used is RBD.
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN
Block no 1 2 3 4 5 6 7 8 9
Store rank X1 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 3 4 8 10 14 17 20 22 25
Treatment group X2 2 5 7 11 13 18 19 23 26 X3 1 6 8 12 15 16 21
24 27
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN (ii) Latin Square Design (LSD) This
allows control of non-interacting extraneous variables in addition
to independent variables. It is similar to RBD technique. Q2. GKK
Ltd. markets a face cream, Hair & Lovely thru 3 types of shops
medical sore, super market, grocery (kirana) store. Sales of the
product vary based on the store type. Sales also vary based on
seasonality, which could be summer, monsoon, winter. The company
desires to test out the impact of 3 different price discounts so as
to rationalize its product pricing all India. For this purpose they
have asked you to conduct an experiment which can be completed in
6-7 months. The process used is LSD.
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN Process : Each extraneous variable
is divided into equal number of blocks/levels. Independent variable
is divided into same number of levels. Extraneous variables make up
rows and columns of the table. IV is assigned to each cell such
that each IV level appears only once in each row/column. Then use
basic design on each cell. 3 x 3 Latin Square:Store Type Time
period 1 2 3 Drug store Price 2 P1 P3 Super market P3 P2 P1 Grocery
store P1 P3 P2
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN (iii) Factorial Design This is used
t measure the effect of 2 or more IV especially if interaction is
likely. Interaction effect: Sum of combined effect Effect of IV1 +
effect of IV2 Q3. Pesti Ltd. desires to launch a new soft drink for
which they have options of (a) 5 levels of sweetness (b) 5 levels
of carbonation. They desire to know the best combination for the
new product. For this purpose, they have conducted an experiment
using factorial design.
RM (Research Methodology)
EXPERIMENTAL RESEARCH DESIGN Process : Each level of 1 IV is
row, and the other is the column. A separate cell is for each
combination. Each combination/cell is rated and average rating
value is put into the cell using basic design). Analysis of
Variance (ANOVA) is used to analyze favorable outcome and
interaction.Carbonation 1 1 2 3 4 5 2 2 4 10 13 2 4 3 6 15 9
Sweetness 3 7 4 8 11 6 4 10 7 5 6 3 5 12 8 5 4 2
RM (Research Methodology)
EXPERIMENTAL ENVIRONMENT :: LAB Design : This is widely used in
package pretesting, ads, products and product concepts (i) Package
testing (projector used): Respondents exposed to package options
Allowed to examine while shopping(thru a controllable screen)
Shopping/identification speed measured (ii) Ads test
(Tachistoscope used): Number of ads screened for visual impact
Designs/messages evaluated fro recall.
RM (Research Methodology)
EXPERIMENTAL ENVIRONMENT (iii) Products tests (blind tests
used): Mainly for taste/utility :: FIELD Experiments Advertising
copy field tests (using Day After Recall) Test marketing :-
Standard Test Marketing (thru manufacturer distribution) -
Controlled Test Marketing (thru marketing research firm) -
Electronic Test Marketing (using electronic test marketing tools) -
Simulated Test Marketing (pre-qualify users)
RM (Research Methodology)
MEASUREMENT :: Definition: Assignment of numbers to
characteristics of objects, persons, states or events, acc to some
rules. :: Concept : A concept is an invented name for an object,
person, state or event.(The above characteristics are called
concept) eg : Sales (measured by volume or by value), Market share
(Own sales / Total sales . Eg: For Bajaj Chetak, market share is
calculated as sales of B.Chetak / total sales of all
motorcycles/scooters. Therefore, be careful when using the term M.S
bcoz the Denominator may vary. So always specify "MS of what??"),
Brand Loyalty (cud be towards product, store, manufacturer)
RM (Research Methodology)
MEASUREMENT Every concept is broken down into conceptual defn
(defines a concept in terms of other concepts) & operational
defn (describes activities to be completed in order to assign a
value to the concept) Eg : take the concept 'Social class'.
Conceptual defn will be "Hierarchical divisions in society to
categories individuals / families with similar values, lifestyle,
interests, behavior" Operational definition will be "Reputational:
Ranking by people who know them Subjective : Based on self ranking
Objective : Based on some / combined chars such as income,
education, occupation"
RM (Research Methodology)
MEASUREMENT :: Measurement Scales : There are 4 types of scales
:(a) NOMINAL Numbers used to characterize objects or events
(numbersused as symbols such as 0 for male & 1 for female)
(b) ORDINAL Numbers/letters/symbols used to rank items (Rank 1,
2 etc.But the problem is that 1st rank may have got 99% and 2nd
rank might have got 88%)
(c) INTERVAL Numerically equal distance on the scale represents
equal distance in property being measured. Location of 'zero point'
not fixed. Eg : body temperature, CGPA scores (d) RATIO Interval
scale with a meaningful zero. Eg : sales (if A sold10 units & B
sold 20, B sold twice as much as A)
RM (Research Methodology)
MEASUREMENT :: Measurement Accuracy : M M C E = C+E measurement
correct value errors(occurs consistently for
Errors could be systematiceach measurement)
or variable
(occurs randomly for each measurement)
RM (Research Methodology)
MEASUREMENT :: Measurement Development : Specify concept domain
Generate item samples Collect data Purify measure Collect data
Assess reliability Assess validity Develop norms
RM (Research Methodology)
MEASUREMENT :: QUESTIONNAIRE DESIGN : Formalized set of
questions for eliciting information A sound questionnaire requires
applying: - Applicable principles - Common sense - Concern for the
respondent - Clear concept of what information is needed - Thorough
pre-testing
RM (Research Methodology)
QUESTIONNAIRE DESIGN
Decision areas involved in questionnaire construction are: (a)
Preliminary decisions (b) Decisions about question content (c)
Decisions about question phrasing (d) Response format (e) Question
sequence (f) Layout of the questionnaire (g) Pretest &
revision
RM (Research Methodology)
QUESTIONNAIRE DESIGN Preliminary decisions The factors that
determine this are - Required information - From which respondents
to get it (whom) - Interview technique cud be direct/indirect,
structured/unstructured, medium personal, mail, telephone,
computer
RM (Research Methodology)
QUESTIONNAIRE DESIGN Question content: For each question, the
researcher must ascertain I. Need for data
II. Ability of the question to produce data III. Ability of
respondent to answer accurately cud be
uninformed/forgetful/inarticulate IV. Willingness of respondent to
answer accurately may not be willing in case of personal
information/embarrassing information/prestige or normative
information V. Potential for external events to bias answer.
RM (Research Methodology)
QUESTIONNAIRE DESIGN I. Need for data The issue is how am I
going to use the data obtained?. The technique is - create
responses - analyse - ensure sufficiency - delete questions not
required
II. Ability of the question to produce data Is one question
adequate or should I break it up. Eg: Do u like big, sporty cars is
ambiguous. It is better to ask Do u like big cars? and Do u like
sporty cars?
RM (Research Methodology)
QUESTIONNAIRE DESIGN III. Ability of respondent to answer
accurately * uninformed (eg: National Bureau for Consumer
Complaints) (eg: what is the current assessed value of your house
can be answerd by the owner of the house. There may be three types
of respondents - a genuine respondent who does not know; A genuine
respondent who knows; Not genuine respondent. To avoid this u can
ask Are u aware of the CAV of your house? Ans: Y/N. What do you
think the CAV is? How close do you think your estimate is to the
actual assessed value? Options: Rs.10000, Rs. 100000, Rs. 200000,
Dont Know * forgetfulness ** omission use aided recall with
condition ** telescoping something that happened in your distant
past, u feel it happened in the near past. Use split interval +
reverse position (split ballot)
RM (Research Methodology)
QUESTIONNAIRE DESIGN **creation use aided recall with condition
* inarticulate use projective techniques like sentence/story
completion, and 2nd level association
RM (Research Methodology)
QUESTIONNAIRE DESIGN IV.) Willingness of respondent to answer
accuratelyPersonal information give explanation Ask at the end Ask
in broad terms Build rapport first Use counterbiasing technique Use
randomized response technique
Embarrassing information -
Prestige & normative technique Ask questions with graceful
exits Eg: reading best-selling book
RM (Research Methodology)
QUESTIONNAIRE DESIGNRandomized Response Technique: Used when
sample size is large (>1000). It should be a large sample
because when applying such probability you should work on large
numbers. Used when information needed is embarrassing. The
assumption is if the respondent is assured that the response cannot
be tracked down to him/her, then he/she would speak the truth. It
is typically used when the incidence of occurrence of certain
events/conditions need to be found out.
RM (Research Methodology)
QUESTIONNAIRE DESIGN Decisions about question phrasing
Researcher & respondent should assign same meaning to
questions. Issues:1. Words used (individual & together) should
be understandable Eg: Would you be interested to buy any of these
products? is different from Which one of these products would you
be interested in buying? 2. Biased words/leading questions : Do you
think India should allow public speeches against democracy? is
different from Do you think India should ban p.s. against
democracy?
RM (Research Methodology)
QUESTIONNAIRE DESIGN Decisions about question phrasing 3.
Clearly state alternatives involved in the questions (clarify): If
there is further shortage of power, do u think there should be a
law to reduce use of A/Cs in establishments in summer? or Do u
oppose such a law? or Should it be left to each establishments
choice? 4. Implied assumptions :Do you favor a ban on ads in movie
halls? Clarify this by asking Would you mind paying extra for this?
5. Frame of reference assumed : Does GIC provide satisfactory
settlement of claims? DO you believe that GIC provides.? Were you
satisfied with GICs settlement of your latest claim?
RM (Research Methodology)
QUESTIONNAIRE DESIGN Response format(i) Open ended Advantages :
No influence on respondents Wide variety of responses Feel for
quality Quotes can be used Disadvantages : Not easy to analyze and
interpret Respondent may not answer adequately Interviewer may not
probe adequately (ii) Close ended multiple choice/dichotomous
RM (Research Methodology)
QUESTIONNAIRE DESIGN Response formatMultiple choice: Advantages
: Easy to administer Reduces interviewer bias Tabulation &
analysis is simple Disadvantages : Development needs a lot of
effort Issues: 1. Number of alternatives (each should appear only
once, all possible alternatives should be included) 2. Alternative
to be balanced 3. Remove position bias by using split-ballot
technique (if alternatives are read out, preference is for last
alternative) (if alternatives are given together, preference is for
first)
RM (Research Methodology)
QUESTIONNAIRE DESIGN Response formatDichotomous: Issues: 1. May
include 3rd option (dont know, no opinion, cant say) 2. Form may
influence results, so use split ballot technique.
RM (Research Methodology)
QUESTIONNAIRE DESIGN Question sequence1. Start with simple,
interesting objective questions 2. Complete questions of one topic
before moving to next topic 3. Move from topic to topic in a
logical way 4. Initial questions to avoid providing biased frame of
reference, or suggest answers to future questions.
RM (Research Methodology)
QUESTIONNAIRE DESIGN Layout of the questionnaire1. Physical
appearance should make it easy to use 2. Layout to minimize
mistakes (avoid go to statements)
RM (Research Methodology)
QUESTIONNAIRE DESIGN Pretest & revision1. Mandatory for all
questionnaires 2. Decisions involved are: - What items to pretest?
- How should it be conducted (std administration,debriefing,
protocol analysis)
- Who should conduct it(Field exec, research associate, research
head)
- Which respondents should be involved(known people, unknown
people, part of sample)
- How many respondents?(Field exec should go to as many
respondents as he can cover in one working day, research associate
should go to as many respondents as required until further
pre-testing does not result in any change in the questionnaire,
research head should pre-test it with 5-8 respondents outside the
office)
3. Language issues: (Especially in India). Do
back-translation
RM (Research Methodology)
QUESTIONNAIRE DESIGN :: Direct Response Attitude ScaleAttitude
is a persons enduring organization of cognitive, affective &
behavioral components & processes with respect to some aspect
of an individuals world. Cognitive Affective Behavioral : persons
beliefs or information about the object : persons feelings of
like/dislike about object : Action tendencies towards object
Substantial part of marketing effort is to assess/influence
attitudes of consumers & intermediaries.
RM (Research Methodology)
QUESTIONNAIRE DESIGN :: Direct Response Attitude ScaleAttitudes
can be measured - directly (DRAS) - indirectly (Derived AS)
Measures attitude directly Two types: - Rating scale (to measure
single dimension) - Attitude scales (to measure multi
dimensions,
it is a combination of rating scales).
Rating scales: Attribute of an object being marked on a
numerically valued continuum or in numerically ordered
category.
RM (Research Methodology)
QUESTIONNAIRE DESIGNTypes of rating scales:Rating Scale
Non-comparative (monadic) Graphic Itemised Graphic Comparative
Itemised Specific Paired comparison Rank order Constant sum
RM (Research Methodology)
QUESTIONNAIRE DESIGNNon-comparative (monadic): Respondent not
provided with a standard to use in the rating. Only one
product/brand is evaluated (hence, monadic). Commonly used at
product concept testing stage in new product development processEg:
Intention to buy, using a Juster scale (11 pt)10.
Certain/practically certain 9. Almost sure 8. Very possible 7.
Probable 6. Good possibility 5. Fairly good possibility 4. Fair
possibility 3. Some possibility 2. Slight possibility 1. Very
slight possibility 0. No chance/almost no chance (99 in 100) (9 in
10) (8 in 10) (7 in 10) (6 in 10) (5 in 10) (4 in 10) (3 in10) (2
in 10) (1 in 10) (1 in 100)
RM (Research Methodology)
QUESTIONNAIRE DESIGNNon-comparative (monadic): Reduce position
bias by reversing order on 50% of questionnaires Top box score is
used for prediction. Top box : % of respondents who mark one of two
most favourable ratings. If at least x% threshold of top box score
is crossed, product concept is taken forward.
RM (Research Methodology)
QUESTIONNAIRE DESIGNGraphic Non-Comparative Rating Scale:
Requires respondent to indicate rating by placing a mark on the
line that runs from one extreme to the attitude being studied to
the other extreme Response is then broken up into categories for
analysis. Never used.
RM (Research Methodology)
QUESTIONNAIRE DESIGNItemized Non-Comparative Rating Scale:
Requires respondent to select one of a limited number of
categories, ordered in terms of their scale positions Types: (i)
D-T scale (Delighted Terrible) scale. Eg: How do u feel after
seeing the movie KKHH? Scale varies as Delighted, Pleased, Mixed,
Unhappy, Terrible. (ii) Percentage scale. Eg: How much did u like
KKHH? Scale varied from 0, 25, 50..100%. 100 is completely
satisfied,0 is not at all satisfied. (iii) Need semantic
differential scale: Eg: To what extent does your AC meet your
needs? Scale is from 1 to 7. 1 is poorly. 7 is extremely well.
RM (Research Methodology)
QUESTIONNAIRE DESIGNRecommendations on using itemized rating
scales: 1. Use precise verbal definitions for at least some
categories 2. Use 5 categories when several scales are to be summed
up for 1 scale 3. Use up to 9 categories when attributes are being
compared across objects by knowledgeable respondents. 4. Use
balanced scales as much as possible. 5. Use odd number of
categories if respondent could feel neutral, even if this is
unlikely 6. Let the choice not be forced upon the respondent. That
is, give option of not knowing/not applicable unless it is known
that all respondents have required knowledge. 7. Adjust rating
scale to nature of information required and characteristics of
respondents.
RM (Research Methodology)
QUESTIONNAIRE DESIGNComparative rating scales: In
non-comparative, rater evaluates the object without direct response
to a specific standard. Hence respondent may apply different
reference points. Eg: rate quality of brand X? - ideal brand -
current brand - average ebrand To obtain uniformity, use known
reference pt (pre-specify). Eg: Compared to the soft drink I
generally drink, Diet Coke is: You have to find out what soft drink
the guy usually drinks (ref pt). Otherwise it is like
non-comparative scale.
RM (Research Methodology)
QUESTIONNAIRE DESIGNComparative rating scales:A A B C D E B C D
E
Visual inspection Brand B is preferred
RM (Research Methodology)
QUESTIONNAIRE DESIGNRank order rating scale :
Requires respondent to rank a set of objects according to some
criterion. All relevant brands need to be included (options).
Simulates actual shopping environment. Faster response time.
Produces only ordinal data.
RM (Research Methodology)
QUESTIONNAIRE DESIGNConstant sum scale :
Requires respondent to divide a constant sum (10 or 100) among
two or more objects/attributes in order to reflect respective
relative preference for each objects/attributes. Can be used for 2
objects (paired comparison) More than 2 : quadric comparison Eg:
Suppose 2 wheeler attributes are to be ranked with 1 being more
important. Following median ranks are produced: Price Economy
Reliability Safety Comfort Style : : : : : : 1 2 3 4 5 6
RM (Research Methodology)
QUESTIONNAIRE DESIGNConstant sum scale :
In constant sum, divide 100 pts among characteristics so that
the division indicates how important each characteristic is in he
purchase decision. Economy: Style: Price: Comfort: Reliability:
Safety: TOTAL = 100
RM (Research Methodology)
QUESTIONNAIRE DESIGNConstant sum scale :
3 groups emerge:Group A Price Eco Rel Saf Comf Style 35 30 20 10
3 2 Group B 20 18 17 16 15 14 Group C 65 9 8 7 6 5
Findings are more detailed than rank order. Can even help to
determine target market. If total > 100 or < 100, adjust it
to 100.
RM (Research Methodology)
QUESTIONNAIRE DESIGNAttitude scales:: scales:
Carefully constructed set s of rating scales designed to measure
individuals attitude towards objects. Response to various scales
may be summed up or examined independently. Scale developed using
method studied before. Forms 1. Semantic differential scale 2.
Stapel scale 3. Likert scale
RM (Research Methodology)
QUESTIONNAIRE DESIGN1. Semantic differential scale Rater is
required to rate objects on a number of itemized 7point rating
scale, bounded by oppositely adjectives at each end (bipolar
adjectives). Eg: for the Zen Fast ----------------- Slow Plain
-------------- Stylish You can also compare 2/more brands on the
same scale: Fast ----------------- Slow Plain --------------
Stylish
RM (Research Methodology)
QUESTIONNAIRE DESIGN1. Semantic differential scale People place
the Zen, Alto, Matiz on different scales. Numbers can be assigned
to each option (1-7) Analysis may be through - Aggregate analysis
(combination of constant sum and SDscale. Multiply the weight by
the rating, for each product)
- Profile analysis
RM (Research Methodology)
QUESTIONNAIRE DESIGN2. Stapel scale: Simplified version of the
SD. Standard is unipolar, 10 interval rating scale with values from
+5 to -5. Any number of values can be used. Values can be all
positive. Eg: What are your views about XYZ stores: Large 3 2 1 -1
-2 Friendly 3 2 1 -1 -2 Low priced 3 2 1 -1 -2 -3 -3 -3
It is good for telephonic interviews, because it saves time. It
is similar to SD, and easy to administer.19
RM (Research Methodology)
QUESTIONNAIRE DESIGN3. Likert scale (summated scale): Respondent
needs to indicate agreement/disagreement to a series of statements
related to objects. Value may be assigned to each option for
analysis. 5 pt Likert 7 pt : Strongly agree, agree, neither,
disagree, strongly disagree : SA, moderately agree, mildly agree,
neither..etc
Eg: XYZ is one of the most attractive stores (SA,A,N,D,SD)
Service at XYZ is not satisfactory (SA,A,N,D,SD) Service is very
important to me (SA,A,N,D,SD)
RM (Research Methodology)
QUESTIONNAIRE DESIGNWhich Scale to Use ?
Depends on: Information requirements of the problem
Characteristic of response Means of administration Cost In general,
- Use multiple measures since summation of measurements is more
accurate.
RM (Research Methodology)
SAMPLING Need for sample planning:Objective of MR is -
estimation - hypothesis testing. Each involves making inferences
about a population on the basis of information from a sample.
Accuracy of survey results are affected by the manner of
sampling.
RM (Research Methodology)
SAMPLING Steps in sample planning:Define Population
Census Vs Sample
Sample Design
Sample Size
Estimation of sampling costs Execute sampling process
RM (Research Methodology)
SAMPLINGPopulation definition:
Definition should include - element - sampling unit - extent -
time Eg: MR for detergents: The population should include price of
each brand kirana outlets, in Mah & Guj, from Jan-Mar 200508-23
Eg: Financial institution with new loan instruments: Population
should be all depositors who have borrowed money from NBFCs, in
metros Mumbai & Chennai, during the last 2 years.
RM (Research Methodology)
SAMPLINGCensus vs Sample:
Census includes the entire population. Sample is a
representative part of the population. Census may not be feasible
all the time. Hence use sample. Advantages of sampling:
speed/timelines & costs However we need to use census in cases
when the population is very small. Eg: If you supply to 15 clients
(the census), you can do a research on all 15 (since the population
is very small). Eg: Medical research for rare disorders.
RM (Research Methodology)
SAMPLINGSample design:
Sample frame: This is a means of representing elements of the
population.Eg: telephone directory. It is not used for
non-probability sampling.
Sampling unit: This is the basic unit containing elements of the
population to be sampled.
Eg: Loan instrument Sampling Unit could be businessmen with over
5 yrs experience (because 5 yrs is the repayment period. So they
would have gone through the entire cycle).
Sampling method: This describes the method by which sample units
are to be selected for measurement. Options: -
probability/non-probabilitysingle unit/cluster of units
stratified/unstratified equal unit probability/unequal unit
probability single stage/multi stage
RM (Research Methodology)
SAMPLINGNon-probability sampling procedures:
(i) Convenience sampling: Based on convenience of sampling.Eg:
to test new food product, add it to menu card at a restaurant.
Advantages: quick, inexpensive
(ii) Judgement sampling/purposive sampling: Representative
sampling is drawn up using judgement selection procedure.Eg: An
organizations new products always failed. Hence sample for study
was competition with compatible products (products that are not
competing, but are being sold in the same way. Eg: shampoo &
toothpaste)
RM (Research Methodology)
SAMPLINGNon-probability sampling procedures:
(iii) Quota sampling: Sizes of subclasses of population
estimated. Sample would contain compatible % of subclasses.Eg: in a
research for fountain pens, subclasses would be students,
employees, businessmen, persons, housewives. Whatever % these
subclasses constitute, you should choose people from these
subclasses in the same % for your sample.
(iv) Snowball sampling: Sample few based on judgment /
random/convenience. Take referrals, include referrals in your
sample. Called Snowball because the size of your sample keeps
increasing as your research proceeds.Eg: MR for travel/tourism,
Readers Digest asks for referrals.
RM (Research Methodology)
SAMPLINGProbability sampling: Sampling unit is selected by
chance. Chance of each unit being selected is known. Designs: (i)
Simple random sampling (SRS): Each sample unit has a known and
equal probability of selection. It is drawn by random procedures
from the sample frame. (ii) Systematic sampling: Each sampling unit
has a known probability of selection. Permissible samples have
known & equal probability. Others have known & zero
probability.Eg: Population size = N, sample required = n, arrange
units (in say, ascending order). At N/n th interval, select a unit
(maybe 1st or 2nd). N = 600, n = 60, Sampling interval = 10. Select
random number between 1-10 (say 5). Then sample elements are
5/15/25/35.
RM (Research Methodology)
SAMPLING(iii) Stratified sampling : Divide population into
strata based on some criteria. Then do SRS within strata.
Difference in: in quota sampling, you use convenience sampling, in
this you use SRS. (iv) Cluster sampling: - Select primary sampling
unit at random. - Sample all elements in selected PSU. - Eg: survey
of sales staff of a pharma company - Select geographical area at
random. - Sample all sales staff in selected area. - Advantage :
Depth interviews possible at low cost. (v) Area sampling: - Select
areas at random. - Within areas, select respondents at random.
RM (Research Methodology)
SAMPLINGSample size :
Sample size determined using following approaches: Unaided
judgment Affordability (all you can afford) Size indicated by
similar studies (take average) Required size per cell Traditional
statistical methods Bayesian statistical model
Unaided judgment: - Arbitrary approach - Precision not clear -
Costing not taken into account - Used commonly, nut should be
avoided
RM (Research Methodology)
SAMPLINGSample size :
Affordability (all you can afford) - Based on budget - At times
practical - Sometimes leads to unnecessary data collection Size
indicated by similar studies (take average) - Can be used as
initial aid - Formal judgment to be used case-to-case Required size
per cell - Used for SRS/stratified random/purposive/quota - Eg:
Fast food joint: Information required from 2 groups
service/businessmen. Four age groups created: 2025/25-35/35-50/50+.
Sample of 30 needed in each group. Hence total sample = 30 * 4 * 2
= 240
RM (Research Methodology)
SAMPLINGSample size :
Traditional statistical methods Sampling distribution of mean:
This is the relative frequency distribution of the means of all
possible samples of size n, taken from a population the size N.Eg:
Super Shoppe has a direct mail campaign during Diwali, which
generated 1250 invoices. Suppose sample size = 50.
Then total possible samples =
1250! (1250 50!) 50!
This is a very high number. Hence (take for eg) 500 samples.
Determine mean of each sample.
RM (Research Methodology)
SAMPLINGValue of mean 38 39.99 40 41.99 42 43.99 44 45.99 46
47.99 48 49.99 50 51.99 52 53.99 54 55.99 56 57.99 58 59.99 60
61.99 62 63.99 TOTAL Frequency of occurrence 1 2 17 39 52 85 110 77
64 37 10 4 2 500 Rel Freq of sample mean .002 .004 .034 .078 .104
.170 .220 .154 .128 .074 .020 .008 .004 1.00
RM (Research Methodology)
SAMPLING These means would be distributed normally in terms of
their relative frequency of occurrences. Relative frequency then
can be a measure of probability. If all possible samples were
taken, we would have obtained mean which would also be distributed
normally with a mean value very close to mean value calculated from
500 samples. Hence sample characteristics can be said to represent
the population
RM (Research Methodology)
SAMPLING
We can infer that a sampling distribution of the mean for SRS
samples has: - Normal distribution - Mean equal to population mean
- SD called standard error of mean . That is equal to population
standard deviation divided by square root of sample size
RM (Research Methodology)
SAMPLING
Basically we are trying to find would want the sample size mean
as possible. The larger to the mean & thus the error N = 1250 N
= 50 N = s2z2 /E2
out the ideal sampling size. We to be as close to the population
the sampling size, the closer it is is smaller.
RM (Research Methodology)
SAMPLINGQ1. A soft drink company desires to determine the annual
per capita consumption of soft drinks in Pune. For this purpose
they would like to carry out a research for which a max error is
0.5lt per capita per year. With a confidence level of 95%. A pilot
sample taken to estimate population dispersion (SD) indicates a
value of 8 liters of per capita per year. Determine the minimum
sample size if the response rate is 25%. S=8 Now the 0.95 Form the
table, the value is read as 1.96.
0.95 / 2 = 0.475.
Therefore, z = 1.96 E = 0.5 N = s2z2 /E2 = 984.
RM (Research Methodology)
SAMPLINGMultiply by 4 for the 25% response rate. Therefore
answer = 984 * 4 = 3936 If the error is 1, and the confidence level
is 60% E=1 Z = 0.84
RM (Research Methodology)
DATA REDUCTIONThis is the process of getting data ready for
analysis. It is intermediate to data processing. Steps involved:
(i) Field controls (ii) Editing (iii) Coding (iv) Transcribing (v)
Generating new variables (vi) Summary statistical calculation
RM (Research Methodology)
DATA REDUCTION(i) Field controlsThese are procedures designed to
minimize errors during actual data collection, thru monitoring and
validation. Monitoring happens when data is being collected.
Validation is checking it after data collection.
(ii) EditingData considered should be current, present ,
readable & accurate. This is the editors responsibility.
Missing data could be filled in through imputation, trend analysis
and plug-in values. For ambiguous data, the editor has to set
standards for evaluation.
(iii) CodingCoding involves establishing categories and
assigning data to them. Categories established as per QD
procedures. Categories shown by numbers as much as possible. Data
assignment is relatively simple for close-ended questions. For
open-ended, create a code-book to indicate system of coding used.
That ensures uniformity.
RM (Research Methodology)
DATA REDUCTION(iv) TranscribingThis is the process of physically
transferring data from measurement instrument to computer storage.
At times may be time consuming, hence it needs to be planned
for.
(v) Generating new variablesThis may be required as a pre-cursor
to analysis. New variables may be generated from combinations of
other variables in data. Eg: persons age/marital status/number and
age of children indicates stage in family life cycle. Data may be
collected thru interval scale and later assigned to classes. Eg:
Family income in Rupees/year may then be put into groups. New
variables may be added thru secondary data Eg: median
income/education in a geographic area. Data may need to be
transformed into another functional form (inverting the 50% where
you have used splitballot technique).
RM (Research Methodology)
DATA REDUCTION(vi) Summary statistical calculation Create basic
data array. Tabulate frequency distributions. - One way tabulation
(with one variable) - Cross tabulation (two variable) - Banners (
more than 2 variables) Calculate central tendency and dispersion
(SD, variance, range) Estimation (either point or interval
estimation). Both are in terms of means & proportions. If the
sample size < 30, use t statistic If the sample size < 30,
use z statistic