《Market Research CourseQuantitative Part.ppt》由会员分享，可在线阅读，更多相关《Market Research CourseQuantitative Part.ppt（43页珍藏版）》请在三一文库上搜索。
1、Overview of Research Methodologies QUANTITATIVE RESEARCH METHODS,Table of Contents,Part I - Introduction To Basic Quantitative Methods Part II - Case Study - Brand Tracking Study Demonstration Using SPSS 14.0 Sample Size Calculation Overview of Data Composition, Scale, Types of variables etc., Explo
2、ratory Data Analysis Basic Statistical Tests To Detect Differences Between groups of doctors Between brands Statistical Tests To Detect Relationships Statistical Hypothesis Testing Part III - Introduction To Advanced Quantitative Methods Factor Analysis Perceptual Mapping Segmentation Analysis,PART
3、IINTRODUCTION TO BASIC QUANTITATIVE METHODS,Definition of Quantitative Research,Quantitative research is numerically oriented and often involves statistical analysis. The main rule with quantitative research is that every respondent is asked the same series of questions. This approach is very struct
4、ured and normally involves large numbers of interviews.,Types of Data,Numerical data generally come in two kinds: MEASUREMENT DATA (sometimes called quantitative data) On an average how many patients with diabetes did you treat in a month? _ patients CATEGORICAL DATA (also known as frequency data or
5、 count data) What is your specialty? Please tick one 1.Cardiologists, 2.Endocrinologists 3.Nephrologists 4.Internal Medicine 5.General Practitioner,Scales of Measurement,Measurement is frequently defined as the assignment of numbers to objects: INTERVAL SCALE in which we can speak of differences bet
6、ween scale points How many years you have been in practice? _years ORDINAL SCALE which orders people, objects or events along some continuum How satisfied your are with brand A? 1.Not at all satisfied 2.Not satisfied 3.Neutral 4.Satisfied 5.Very satisfied NOMINAL SCALE is not really a scale as they
7、do not scale items along any dimension What is your gender? Please tick one 1.Male 2.Female In which state do you normally practice? Please tick one 1. ACT 2.NSW 3.NT 4. QLD 5. SA 6.TAS 7.VIC 8.WA,Composition of Data,A data typically consists of: VARIABLES are properties or characteristics of a stud
8、y or therapy type etc., e.g. attitudinal statements, product characteristics, functional and emotional attributes, patient flow etc., CASES typically correspond to respondents e.g. Doctors, Patients, Pharmacists etc., RESPONSES are values assigned to each variable e.g. 1, 2 or 3 etc.,Types of Variab
9、les,The variables fall under two types depending on different values assigned: DISCRETE VARIABLES in which the variable can take on only one of a relatively few possible values e.g. Gender, Location, Specialty etc., CONTINUOUS VARIABLES in which the variable could assume any value between the lowest
10、 and highest points on the scale e.g. Number of patients treated, Diagnostic parameters like BP, LDL-C, HDL-C, HbA1c etc., We can also distinguish between different types of variables in an additional way: DEPENDENT VARIABLE is the variable of interest or whose values we wish to predict e.g. Overall
11、 brand satisfaction INDEPENDENT VARIABLES (also referred as explanatory variables or predictor variables) are the variables which influences or affects dependent variable. e.g. Importance statements, product characteristics,INFERENTIAL Inferential statistics are concerned with using SAMPLES to infer
12、 something about POPULATIONS. Population can range from a relatively small set of numbers, Which is easily collected, to an infinitely large set of numbers, which can never be completely collected. Hence we are forced to draw only a SAMPLE of observations from that POPULATION Range of statistical te
13、sts are available to choose from which depends on the objectives of the study,Types of Data Analysis,EXPLORATORY Exploratory techniques are used for presenting data in visually meaningful ways. Also, these techniques are used for initial data screening purposes. Frequency distributions Histograms Bo
14、x and Whisker plots Stem and Leaf plots Missing values Outliers,DESCRIPTIVE Descriptive statistical tests are used to describe a set of data. Mean Median Mode Sum Measures of dispersion Minimum Maximum Range Variance Standard deviation Standard error Distribution Kurtosis Skewness,Probability Sampli
15、ng Methods This method gives all members of the population a known chance of being selected for inclusion in the sample,Methods of Sampling,Non-probability Sampling Methods The quotas (quota sampling) within subgroups are set beforehand usually proportions are set to match known population distribut
16、ions.Interviewers then select respondents according to these criteria rather than at random. The subjective nature of this selection means that only about a proportion of the population has a chance of being selected in a typical quota sampling strategy,Sample Size CalculationGeneral Rules,What marg
17、in of error can you accept? The margin of error is the amount of error that you can tolerate. Lower margin of error requires a larger sample size 5% is a common choice What confidence level do you need? The confidence level is the amount of uncertainty you can tolerate. Higher confidence level requi
18、res a larger sample size. Typical choices are 90%, 95%, or 99% What is the population size? How many people are there to choose your random sample from? What is the response distribution? For each question, what do you expect the results will be? If the sample is skewed highly to one end, the popula
19、tion probably is, too. If you dont know, use 50%. This gives the largest sample size. The most conservative choice is 50%,Sample Size Calculation Infinite vs. Finite Population,INFINITE POPULATION The sample size does not change much for populations larger than 20,000 FINITE POPULATION In pharmaceut
20、ical market research studies we commonly encounter a small population (i.e. finite population) in certain therapy areas such as Oncology, Nephrology, Neurology etc., In such situations the sample size calculation must incorporate FINITE POPULATION CORRECTION (FPC) With small populations, we may have
21、 to interview a significant proportion of the population to achieve stable estimates However, once a population reaches about 5000 or more, we can generally ignore the finite population correction factor, as it has very small impact on sample size decisions,Sampling Design Sources of Error,BIAS: occ
22、urs when flawed sampling procedure is used. Bias cannot be measured, and the selection of a larger sample does not correct for the fact that the sample is biased SAMPLING ERROR: occurs when samples of respondents deviate from the underlying population. If we have used random sampling method, any sam
23、pling error is due to a chance. With random sampling, we reduce sampling error by increasing the sample size. MEASUREMENT ERROR: arises from many sources data collection methods interviewers quality and experience respondents understanding of the questionnaire type of scale questionnaire design cont
24、ext of data collection,Statistical TestingParametric vs. Non-Parametric Tests,The selection between a parametric or non-parametric tests depend on whether the data being analysed are sampled from a population that are NORMALLY DISTRIBUTED (also referred as Gaussian distribution) or not PARAMETRIC ST
25、ATISTICAL TESTS: are used when normality is met NON-PARAMETRIC STATISTICAL TESTS (or distribution free tests): are used when normality is not met A normal distribution is a symmetric, unimodal distribution, and “bell shaped”,Statistical TestingProperties of Normal Distribution,68.2% of the responses
26、 fall within a standard deviation of +1 The mean, median and mode are equal Generally the distribution tend to become normal for a sample size of greater than 30,Most marketing questions (i.e. statistical) fall roughly into one of two overlapping categories: DIFFERENCES Is there any difference betwe
27、en endocrinologists and GPs in terms of overall satisfaction of diabetes treatment? Is there any difference between brands? RELATIONSHIPS Is there any relationship between importance of product attributes and overall performance of a given brand?,Statistical TestingDifferences vs. Relationships,Stat
28、istical Hypothesis Testing,A hypothesis test usually derives from a prior research hypothesis (usually QUALITATIVE stage) stating that a relationship exists between two or more variables, or two or more groups will be different on some characteristics. For example, Hypothesis: Among doctors, orthopa
29、edics believe In post-menopausal women with osteoporosis, Vitamin D inadequacy is widespread and needs to be corrected” than GPs To test this research hypothesis statistically, we would formulate two statistical hypothesis: Null hypothesis: the mean for orthopaedics on the above statement is equal t
30、o or less than the mean for GPs. Expressed in symbols: Ho:ortho GP Alternative hypothesis: the mean for orthopaedics will be greater than the mean for GPs. Expressed symbolically: H1:ortho GP The two hypotheses are mutually exclusive. They cannot both be true. As a researcher performing a statistica
31、l test, we must ultimately make one of two decisions relative to the null hypothesis Choosing a statistical test to test the research hypothesis One-tailed test: is used if we know the direction of the hypothesised difference i.e. Orthopedics believe than GPs Two-tailed test: is used if we do not kn
32、ow the direction of that difference,Choosing a Basic Statistical Test,DIFFERENCES,RELATIONSHIP,NOTE: * These analyses will be demonstrated with the help of a case study,PART IICase Study - Brand Tracking Study,Demonstration Using SPSS 14.0 Sample Size Calculation Overview of Data Composition, Scale,
33、 Types of variables Exploratory Data Analysis Basic Statistical Tests To Detect Differences Between groups of doctors Between brands Statistical Tests To Detect Relationships Hypothesis Testing,PART IIIINTRODUCTION TO ADVANCED QUANTITATIVE METHODS(MULTIVARIATE TECHNIQUES),Case Study Factor Analysis
34、Perceptual Mapping Segmentation Analysis,Multivariate Techniques,Multivariate techniques are broadly grouped into two categories: INTERDEPENDENT techniques are used to either grouping similar customers or variables To identify target customer segments (grouping customers) To identify underlying atti
35、tudinal / emotional “themes” or “factors” (grouping variables) DEPENDENT techniques are used to study the effect of a number of variables on one or more variables To identify drivers of brand satisfaction (satisfaction vs. product, attitudinal, emotional attributes),Choosing a Multivariate Technique
36、Interdependent Techniques,NOTE: * These analyses will be discussed in detail,Choosing a Multivariate TechniqueDependent Techniques,Factor Analysis,Factor analysis is an exploratory data analysis technique often used as a data reduction method Factor analysis transforms a set of correlated variables
37、into fewer conceptual dimensions known as “factors” Factor names are provided by analyst, based on dominant or common attributes within the bundle of attributes,Sample size considerations The common rule is to have at least 10-15 responses per attribute,Factor AnalysisCase Study,MARKETING APPLICATIO
38、N A marketer have measured 30 different PERCEPTION ATTRIBUTES that are considered important to a specific therapy area The marketer wanted to identify underlying themes or factors within those 30 attributes rather than evaluating individual attributes. TOOL Based on intercorrelations among these 30
39、ATTRIBUTES,factor analysis reduced them to 7 THEMES OR FACTORS This has helped the marketer to focus and develop marketing strategy accordingly This has also helped the market researcher to reduce the number of attributes to be incorporated in the questionnaire for future studies,Factor AnalysisCase
40、 Study,QUESTIONNAIRE Please indicate your level of agreement with each of the following statements as they apply to the treatment of a “specific therapy area”. Using the 5-point scale shown, where “1” means “strongly disagree”, “5” means “strongly agree”, please cross the most appropriate response.
41、Attribute 1 Attribute 2 . . . Attribute 30,Factor AnalysisIdentification of Underlying Themes,The 30 perception statements rated by respondents are simplified into 7 key themes or factors using factor analysis*.,Factor Analysis,Factor Analysis,The 30 perception statements rated by respondents are si
42、mplified into 7 key factors or components using factor analysis*.,Factor AnalysisIdentification of Underlying Themes,Perceptual Mapping,Perceptual map is a visual representation of TWO data sets such as different brands and product attributes different brands and demographic traits different corpora
43、tions and key image attributes etc., Perceptual map is produced by using TWO different techniques: Correspondence analysis Multidimensional scaling,Perceptual MappingCase Study,MARKETING APPLICATION A marketer wanted to identify how 5 DIFFERENT BRANDS are perceived on 8 KEY ATTRIBUTES as applied to
44、a specific therapy area TOOL Correspondence analysis produced a perceptual map which positions brands against attributes on a two dimensional map This has helped the marketer to UNDERSTAND BRAND POSITIONING,Perceptual MappingCase Study,QUESTIONNAIRE Based on your knowledge and experience with the br
45、and, please identify presence or absence of below 8 attributes that a brand may have Brand 1 Brand 2 Brand 3 Brand 4 Brand 5 Attribute 1 Attribute 2 . . . Attribute 8,Perceptual Mapping: Brands vs. Attributes,Brand E,Brand A,Brand B,Brand D,Brand C,Works fast,Improves bone quality,Prevents hip fract
46、ures,Safe for long-term use,Convenience,Prevents vertebral fractures,Well tolerated,The relative perceived positioning of the various 5 brands vs. 8 product attributes,Increases BMD,Segmentation Analysis,The purpose of segmentation analysis is to form groups of respondents (doctors, patients etc.,)
47、such that, with respect to clustering variable (s), each group is as homogenous as possible and as different from the other groups as possible. There are TWO main routes to segmentation: A priori segmentation (referred as OUTSIDE-IN) The segments are ASSUMED before-hand such as based on demographics
48、, geographic etc., Cluster-based segmentation* (INSIDE-OUT) The segments are NOT ASSUMED before-hand but identified based on homogeneity exhibited by respondents on certain parameters such as behavioural attributes, perception statements, attitude towards new drug introduction etc.,NOTE: * This will
49、 be discussed in detail,VARIABLE SELECTION is the single most crucial step in cluster analysis The selection of variable depends on the type of business objective that we have in mind. We may want to segment respondents based on “benefit segmentation,” or “need-based segmentation” or “attitudinal se
50、gmentation” etc., The variables that meet our business objective are chosen for cluster analysis The segments are then profiled to identify distinct differences between the segments The number of clusters (i.e. segments) depends on the following characteristics: Meaningfulness How sensible or meanin