UO Marketing 390 Midterm
Nominal
- most basic, only categories (no order) - property: identity - Univariate Analysis: Frequency, Mode -EX: Did you earn any income last year? Yes or No?
Interviews Pros
- not influenced by others - great depth, follow-up - candid, sensitive topics - expensive - time consuming
Open-Ended Questions
- only the question - EX: What would you like to learn from this class? (leave a box for a response)
Questions for Focus Groups
- open ended (what did you think of X?) - not quantitative - Follow ups: ask about attributes and/or influences (no "why" questions) - Engaging
Demographics
- primary data - age/gender/ethnicity/income
Attitudes and Beliefs
- primary data Belief - patter of knowledge that you hold as true - FACT BASED (or what you think is fact) - difficult to change Attitudes - overall evaluation of something - OPINION BASED - easier to change - common attitude scales *changing beliefs/attitudes done through promotion
Industry: Direct Competitors
- produce a very similar product offered in the same market - similar product functions, attributes, benefits, and price - EX: Honda/Toyota/Nissan
Longitudinal Analysis Survey (descriptive research)
- repeated measures over time of a fixed sample (same people over time) - longitudinal sample is called a "panel" of participants
Interval
- show DISTANCE between points - property: identity, order, AND distance - Univariate Analysis: Mode, Median, Mean, Standard Deviation -EX: What is your annual income? (list of seven ranges of answers)
Closed Ended Questions
- show questions AND choices Dichotomous (two options) Multiple Choice (3+ options) Scales (continuum) - have fixed alternatives Dichotomous Questions = 2 alternatives - Do you think that gasoline will be more expensive or less expensive next year than it is now? Answer: More or Less Questions with more alternatives - are the answer choices mutually exclusive or collectively exhaustive
Ratio
- shows distance between points and has a natural zero - property: identity, order, distance, AND natural zero - Univariate Anaylsis: Mode, Median, Mean, Standard Deviation -EX: How much do you earn annually? (answer is fill in the blank)
Cross-Sectional Study Survey (descriptive research)
- single point in time - measures of a sample selected from a population
True Experiment
- the best way to test for causation (controls for confounding variables)
External Environment (secondary data)
- things outside the company - opportunities + threats from SWOT - Industry (competitors, market share distribution, overall growth, etc) - PEST: Political, Economic, Sociocultural, and Technological Factors
Internal Environment (secondary data)
- things within the company - strengths + weaknesses from SWOT
Secondary Data
- use data to understand the marketing environment Internal - company financials - company strengths and weaknesses External - industry and competitors - political factors - economic factors - socio-cultural factors - technological factors
PEST Factors
- useful frame work for understanding macro-environment trends Political/legal Factors - laws, regulations Economic Factors - average income, purchasing power, inflation Socio-cultural Factors - values, lifestyles, age-groups Technological Factors - innovation, infrastructure
Univariate - Frequency Counts
A count of the number of cases that fall into each of the response categories Used mostly with categorical measures Graphed by pie, bar, histogram (if continuous)
Industry Indirect Competitors
Can be classified two ways - produce a very similar product as part of a wider range of products (liquor store/safeway) - produce a different product that can serve as a substitute (amtrak/southwest)
Univariate - Central Tendency and Dispersion
Central Tendency - Mode: most frequent response - Median: response that splits dataset in half - Mean: (xbar): arithmetic average Dispersion (only for continuous variables) - Range (min-max) - Standard Deviation (s): average distance from the mean - Variance (s^2)
REVIEW: Graphing Multivariate Analysis
Comparing Groups (cross-tabs, t-tests, ANOVA, etc) - bar graphs Correlation - scatterplot Regression - scatterplot with fit line
Open-ended vs Closed
Considerations - comparability of data - richness of data Criteria - for surveys, use closed response whenever possible - if open-ended question are necessary, give respondents a frame of reference
Rating Scale - Likert-type
Considered "Quasi-Likert" Used when you don't need all the labels, but want more than just the end-points labeled (Dislike-Neutral-Like)
Sampling Procedures: Non probability - Convenience
Convenience Sample - elements are sampled because they are in the right place at the right time (Just the ones we can reach) Usually ALL members of sampling frame are contacted and sample consists of those who decide to respond Considerations: - easiest to execute, results in largest sample size - not necessarily representative of the population - to make the best of it maximize the sample size, aim for diversity in recruiting, make sure everyone in sample meets criteria for target population, design a high-quality survey, and acknowledge the limitations of this procedure in your report
REVIEW: Why Regression instead of Correlation?
Regression allows you to control for other possible predictors Regression provides a line of best fit - allows you to predict exact values of Y given specific values of X Regression provides a more meaningful metric (R^2) You can also use dichotomous (0/1) predictors in regression - cannot use categorical predictors with 3+ levels - use dummy-coding or contrast-coding if necessary
Calculating Response Rates
In general: RR = CS / E - Response Rate - Completed Surveys - Number of eligible responding units in the sample
Population Standard Deviation Unknown
Option 1: Previous study Option 2: Estimate by using range of variation - For a noramlly distributed variable, the range is approximately equal to + or - 3 standard deviations - sigma = range / 6
Sampling Terminology
Population - the total of all the elements that share some common set of characteristics Sample - a subgroup of the population selected by the researcher Element - individual object (or person) about which the information is desired Sampling - the process of selecting a sample of elements from a population of interest
Variation
Population size doesn't matter, its the VARIATION of the population Low variation = small sample size is representative High variation = large sample size
Precision (H)
Precision (H) = Confidence Interval Width / 2 If confidence interval is between 1.45 and 18.54, then: - CI Range = 18.54-1.45 = 17.09 - Therefore, H = 17.09/2 Sample ACF's estimate of mean age (xbar = 10) is within a MARGIN OF ERROR of +/- 8.55 years
*Information Sought*
Research Problems lead to Variables and Hypotheses (guided by Focus Groups/Interviews Results) lead to Information to be collected in questionnaire
Field Experiment
Research study in a REALISTIC SITUATION in which an independent variable is manipulated under as carefully controlled conditions as the situation will permit - includes a situation with neutral conditions - control of variables other than X or Y is challenging - X variable are manipulated
Lab Experiment
Researcher has complete control over environment and manipulation of independent variable - situation with exact conditions - control of variables other than X or Y is maximized - X variables are manipulated
Non-Sampling Error
Researcher mistakes Biases (both instrumentation and response biases
Response Bias - Social Desirability
Respondents are prone to reports what's socially acceptable Most common and powerful bias against accurate responding Get around this bias through: - the "average other" - projective techniques - disguised scales - SD scale
Constructing a "Confidence Interval"
Sample mean is an estimator of the population mean - construct confidence interval around sample mean to reflect how accurate the estimate is
Why is sampling done?
Sampling is done because we may generalize results to the population - sample is drawn from the population, sample statistics allow inferences about population parameters
Cross-Tabs
Shows relationship between two categorical variables - both variables must be categorical (nominal or ordinal) Purpose: See if one variable (predictor) influence another (outcome) Create a matrix of frequencies - Rows: levels of predictor - Columns: levels of outcome - Cells: frequencies (percentages) in each crossing of categories Compute percentages in the direction of the predictor (X) variable
Communication Methods
Projective Techniques Focus Groups/In-depth interviews Surveys
Disguised Methods
Projective Techniques - word association - sentence completion - story telling Us vs Them Method - undisguised: are you afraid to fly? - disguised 1: do you think other people are afraid to fly? - disguised 1: why do you think some people are afraid to fly?
Rating Scale - Sliders
Provides most variation in responses - good for detecting subtle differences between respondents More interactive, engaging for respondents
Instrumentation Bias - Leading Questions
Questions that lead respondents to a particular answer
Sampling Procedures: Non probability - Quota
Quota Sample - sample chosen so that the proportion of sample with a given attribute equals proportion of population with respect to the same attribute (proportion of colors in sample matches the proportion of colors in the jar) Considerations: - more work to execute but more representative of population - requires constant checking on incoming data and adjustments to sampling - guarantees diversity on dimensions important to the researchers - requires relatively large sampling frame
Step 5 - Re-examine, Pretest, and Revise
Re-examine steps 1-7 and Revise if Necessary
Disguise in Primary Data Collection
Undisguised - participants know everything about the true research purpose Disguised - little or no knowledge of the true research purpose Why use disguise? - respondents may be unable or unwilling to answer honestly
Step 3 - Content of Questions
Use a variety of levels of measurement (NOIR) Use (almost) exclusively closed questions Develop questions that directly address your hypotheses Always collect basic demographic data Avoid extra, unnecessary questions (minimize survey length) Rules of wording
Step 1 - Specify Information Sought
clearly define your research problems specify variables/constructs to study hypothesize relationships between key variables
Research Problems
identify what the researcher needs to research
Instrumentation Bias
induced by instructions, questions, scales, or response options can be eliminated types: - ambiguity - double-barreled - leading question
How to sample when population variation is high
larger sample sizes more diversity (ie variation) in sample random sampling procedure if possible - because mean and std. dev. are unbiased estimators of population parameters
Step 2 - Method of Administration
mail, telephone, in-person, email or web- based
How to tell if there's a lot of variation in the population
shape of distribution (wide = high variation) standard deviation (sigma) (high value = high variation) high ration of sigma/range (above 33% is high)
Interpreting Standard Deviation
sqrt(sigma^2) = population standard deviation Is 5.61 high or low? Range = 18-3 = 15 Generally about 2/3 of the population fall within one sigma (std. dev.) of the mean Can also quantify as st. dev / range ratio: 5.61/15 = 37% ANYTHING OVER APPROXIMATELY 33% is fairly high
Rating Scale - Likert
statement to be evaluated 5 or 7-point scales are the norm each response option has a label
Decision Problem
the clients view of the problem
Primary Data
- information collected specifically for the PURPOSE OF THE PRESENT STUDY
Projective Techniques
- involves the use of an AMBIGUOUS stimulus that an individual is asked to: describe react to expand upon build a story around - can be incorporated into surveys, interviews, or focus groups
Interviewing Basics
- 1 participant / 1 interviewer at a time - interviewees as representatives - data = transcript of interview
Focus Group Basics
- 5-10 people at one time (6-8 preferred) - homogenous groups - provide incentive for participation - multiple groups to get heterogeneity - comfy, circle seating, audio/vidoe recorded - snacks or drinks
Ordinal
- Categories with order - property: identity AND order - Univariate Analysis: Frequency, Mode, Median - Use dummy-coding for multivariate anaysis -EX: Do you earn more or less than $50,000/year? More/Less
Three Necessary Conditions to show Causation "to show that X (independent variable) CAUSES Y (dependent variable"
- Concomitant Variation (significant relationship between X and Y - Time order of occurrence (X occurs before Y) - Nonspurious effects (rule out alternative explanations
Secondary Data
- Existing information initially collected for some OTHER PURPOSE
Flow of a Focus Group
- Introduction - Prepared Questions - Recognize and exert MILD control of different participant types - Transition smoothly between topics - Start general question, get more specific throughout - Three-step conclusion
Basics of Experimental Design
- Must have at least two groups of participants - Two groups should be as closely matched on everything else as possible - Participants must be randomly assigned to groups - Observe outcome on both groups
Types of Projective Techniques
- Word Association - First Reaction - Sentence Completion - Storytelling
Industry
- aggregate of buyers and sellers around a product offering - competitors - market share distribution - etc
Flow of an Interview
- agree to a game plan - ask questions about the person - ask questions about the research problem - make concluding remarks
Strategy-Oriented Decision Problems
- asks about planned changes in response to the unplanned change - stated as the question: HOW can we strategically address this unplanned change
Motives
- based on consumer needs - people need outcomes, not products Basic Classification - Functions needs (basic or extra): practical outcomes of consumption - Social needs: outcomes related to self-image and social relationships
Exploratory Research
- cast a wide net to examine new questions and generate hypotheses - done through QUALITATIVE methods (projective techniques, focus groups, in-depth interviews) - purpose is to develop HYPOTHESIS
Discovery-Oriented Decision Problems
- common with unplanned changes - stated as the question: WHY is the unplanned change happening?
Descriptive Research
- current state of the environment - determine relationships between variables and tests hypotheses - usually done through QUANTITATIVE METHODS - SURVEYS= primary - DATABASES = secondary
Applications of Descriptive Research
- describe a population - determine proportion of people who behave in a certain way - make specific predictions - determine relationships between variables
Fundamental Question of Causal Research
- did X cause Y? - X and Y may be related - does something about X CAUSE a change in Y? - does X influence some other variable (Z), which in turn causes a change in Y?
Focus Groups/ Interviews Pros
- flexible rich qualitative data - gets at the WHY - generates ideas
Comparative Scale - Paired Comparison
- good in a telephone survey - overwhelming in written/online surveys -EX: For each of the following pairs, which soft-drink do you thin is better (please check on soft-drink for each pair)
Comparative Scale - Rank-Order
- good in written/online survey - hard to administer on the phone -EX: rank the following soft-drinks from 1 (best) to 5 (worst) according to your taste preference
Focus Groups Pros
- group dynamics encourage creativity - relatively inexpensive - efficient
Comparative Scale - Constant Sum
- hard in telephone surveys - better in written/online surveys - allows for finer discrimination - confusing with too many objects (limit 4-5 max) -EX: Allocate a total of 100 points among the following sodas depending on how favorably you feel toward each
Showing Causation Example
- independent variable = SMOKING - dependent variable = LUNG CANCER - exploratory stage = don't know what (if any) relationship exists - descriptive stage = establish the relationship - causal stage = manipulate IV and observe the effect on DV
Procedure for Developing a Questionnaire
1 - Specify Information Sought 2 - Determine Method of Administration 3 - Determine Content of Individual Questions 4 - Determine Question Sequence 5 - Re-examine, pretest, and revise
Awareness/Knowledge
1) Advertising recognition/recall Recognition = recognition cue - do you remember seeing this ad? Aided recall = category cue - what ads for computers have you seen lately? Unaided recall = no cue - what televised ads have you seen lately? - THE BEST 2) General - Qualitative: Tell me what you know about X (open-ended) -Quantitative: How familiar are you with X (continuous scale)
Specifics on Sample Size
30 is a rule of thumb for MINIMUM sample size - sampling distribution becomes normal at n=30 A more exact sample size (n) accounts for three factors - Variation in population (St. Dev.) (Sigma) - Desired Precision (H) - Desired Confidence (Z)
Commonly used Confidence Levels
90% - Z = 1.65 95% - Z = 1.96 99% - Z = 2.58
Multivariate Analysis - Regression
A more detailed, sophisticated correlation analysis - Both X and Y are continuous measures - Allows you to test the effects of multiple Xs on Y at the same time - Provides a statistic (R^2) that shows % of variance in Y explained by X Fits a line through a scatter plot - Provides a measure of "fit" of the line to the dots - Interpret this "line of best fit" ** error is very low = fit is very high ** a little more error = slightly worse fit
Conventional Scale Types
A scale is a numerical continuum along which respondents place themselves. Used to measure attitudes, perceptions, preferences, and other unobservables Rating Scales - Likert (and likert-type) - Semantic Differential - Sliders Comparative Scales - Paired comparison - rank-order - constant sum
Characteristics of all Markets
All markets have some things in common: - composed of people - these people have wants/needs that can be satisfied by "products" -these people have the ability and willingness to exchange resources for them
Instrumentation Bias - Ambiguity
Ambiguous words and questions
Error
Anything that causes a skewed result
Relationship between Confidence and Precision
As confidence gets higher, precision gets higher
Defining a Market in our Context
As marketers, we often define a market as the overall group of people seeking a certain product category Often qualified by other descriptives to narrow it down
Relationship between Sample Size and Precision
As sample size gets larger, precision get higher
Rating Scale - Semantic Differential
Assesses bi-dimensional intensity 7 to 9-point scales are the norm Responses labeled with opposite adjectives/phrases (shy/outgoing and unorganized/organized)
Consideration in Scale Development
Balance - same number of positive/negative response options Forced - existence of a true midpoint - use forced when you don't want neutral responses Number of scale points (rating scales) - for most rating scales use 7 +/- 2 Number of items per construct - item: a single question or statement that requires a response - scale: a collection of one or several items used to measure something - construct: the idea or object being measured
Univariate Analyses (ie. Descriptive Statistics)
DESCRIBES the sample, one variable at a time Possible with categorical measures: - frequency counts - mode - median Possible with continuous measures: - frequency counts - mode - median - mean - standard deviation - range
Desired Precision and Confidence Level
Degree of Confidence -Confidence DECREASES as sample size increases Degree of Precision - As need for precision INCREASES, larger samples are required There is always a trade-off between confidence and precision - Higher precision means lower confidence and vice versa
Common Types of Primary Data
Demographics Personality/Lifestyle Attitude and Beliefs Awareness/Knowledge Intentions Motives (based on consumer needs)
ANOVA
Determine whether 3+ groups differ on some continuous measure - if so, which groups differ from each other
Independent Samples t-test
Determine whether TWO groups differ on some continuous measure
Observational Methods
Direct observation Neielsen TV Ratings Retail Scanner Data Online tracking
Response Bias
Due to the mentality or predisposition of the respondents Can only be reduced not eliminated types: - social desirability - acquiescence - item/option order
Sampling Error- Coverage
Failure to include part of the target population in the sampling frame - sampling frame is not representative Only contributes to the "error" if there is something systematic about those who are not in the sampling frame - is there something that sets these people apart? - is this shared trait relevant to your research?
Sampling Error - Response Rate
Failure to reach entirety of planned sample - subjects who receive your survey don't answer it Think of it as proportion of planned sample that actually responds Only contributes to "error" if there is something systematic about those who do not respond
Five Major Bases for Segmentation
Geographics - geographic location Demographics - age, gender, income, etc Psychographics - lifestyle, personality, etc Benefits Sought - needs satisfied by product Usage-Rate - frequency of use or purchase *some bases are composed of several different variables * can segment based on one or multiple variables or bases
Relationship: Sample Size, Confidence, and Precision
Given confidence level, as sample size (n) increases, precision increases (CI width becomes narrower) Given precision, as sample size (n) increases, we become more confident about our estimate
Sample Variance
UNBIASED The sample variance is an unbiased estimator - on average, the sample variance equals the population variance
Step 4 - Question Sequence
In the beginning of the survey: - start with simple, interesting, and nonthreatening questions - questions that ask for opinions are often good Use the Funnel Approach - start broad and progressively narrow down the scope First - Collect Responses relevant to the purpose of the study (attitudes, intentions, perceptions) Last - Collect responses relevant for classification (demographics, psychographics, etc) Place difficult of sensitive questions later - don't ask unless absolutely necessary - if the respondent were to stop once sensitive questions are reached, at least they stop after the purpose of the study has been addressed Order: - Relevant to study to demographics/psychographics - Interesting, non-threatening to difficult/sensitive - EX: Funnel Approach: Broad - which three UO building come to mind first? Narrow - What is your opinion of Lillis Business Complex?
Multivariate Analysis - Comparing Group Means
Independent Samples t-test - compare mean score on same outcome variable between 2 separate groups ANOVA - compare mean score on same outcome variable between 3 or more groups
Sampling Error
Information lost due to studying a sample rather than the whole population Makes sample less representative of the population Difference between sample results and results that would have been obtained had you studied every member of the population (ie. Margin of Error) - Wider margin > wider confidence interval > more sampling error > less precision - can be reduced by increasing sample size
How much Precision is Needed?
Judgement call specified by the researchers Precision is sometimes called "margin of error"
Identify the Sampling Frame
List of target population elements from which a sample (n) will be drawn (Everyone from target population who is listed on some kind of record) We do not always know all the elements in the population Commonly used sampling frames include: - customer databases - telephone directories - online message boards or groups
Experiments in Marketing
Manipulations (IVs) usually involve some form of the following: - versions of an advertisement - scenarios involving companies or other consumers - closely matched product features or designs Typical Outcomes (DVs): - attitudes toward ad, brand, or product - purchase intentions/behavior - memory of an ad
Converting Continuous into Categorical Measures
Measures at higher levels have all properties of measures at lower levels Why do this? - EASE OF INTERPRETATION FOR MANAGERS Median Split: Split respondents into two groups - low and high (below or above median) - used primarily in multivariate analyses Two-box Technique: The percentage of respondents choosing of the the top two positions on a rating scale
Multivariate Analysis - Correlation
Measures strength of relationship between two CONTINUOUS variables (r) Bidirectional - can NOT say that one causes or influences the other - only that the two variables are associated Can only take values between 0 and 1
Non-Sampling Error - Office Errors
Mistakes made by researchers and data collectors Errors that arise when coding, tabulating, or analyzing the data Office errors can be located by: - examining frequency distributions on all variables - checking a sample of questionnaires against the data file
Coding Restrictions
Nominal - just about anything goes - code each category in any order you want (1 = yes, 0 = no) Ordinal - codes must maintain order - code each category in rank order Interval - codes must maintain order and relative magnitude - code the scale from 1 to 7 (if 7 scale points) Ratio - coding is generally not necessary - enter response as it is unless you need to convert the scale
Select a Sampling Procedure
Nonprobability Sample - cannot calculate the probability of each element being selected - inferences are limited to the sample, cannot be made about the population - thus, results ARE NOT generalizable from the sample to the population Probability Sample - one can calculate the probability of each element being selected - inferences can be made about the population, and not just the sample size - thus, results ARE generalizable from the sample to the population
Sampling Procedures: Probability - Simple Random
Simple Random Sample - each element as the same chance of being selected - completely random (marbles are rotated and chosen blindly) Passing out surveys to unknown people "at random" is random in the everyday sense, but NOT RANDOM IN A SCIENTIFIC SENSE Considerations: - best sampling procedure because it is reliable and produces a sample that is representative of the population - requires access to a large and accurate sampling frame - need a high response rate for it to work (ideally 100%)
Response Bias - Acquiescence
Some people have a tendency to agree to almost any question The remedy: include reverse-coded items in scales Consider if your survey has a lot of similar scale items in a row
Sampling Procedures: Probability - Stratified
Stratified Sampling - Divide population into mutually exclusive and collectively exhaustive subsets (strata) - A simple random sample of elements is chosen independently from each stratum (divide jar into several "subjars" and draw a simple random sample from each) How do you make the strata? - Homogenous groups - mutually exclusive and collectively exhaustive
Is it "Statistically Significant?"
Test against the common cutoff: p < .05 - corresponds to 95% confidence Need to do a Chi-square test
Multivariate Analyses (ie. Inferential Statistics)
Tests relationships between tow or more variables - allows you to INFER what this relationship is in the population - provides a value (p) so we know whether this relationship is statistically significant Independent Variable - the predictor (X) Dependent Variable - the outcome (Y) Basic question: do changes in X predict changes in Y?
Instrumentation Bias - Double-Barreled
Two (or more) questions in one
Significance Testing in Regression
Two indicators - Fit should be significantly larger than error = F-test -Slope (b1) should be significantly different from zero - b1 = change in Y given a 1-unit change in x = t-test
Types of Multivariate Analysis
Types of analysis depends on nature of X and Y X and Y are both categorical - CROSS-TABULATION ("cross-tabs") X is categorical and Y is continuous - INDEPENDENT SAMPLE T-TESTS (two levels of X) - ANOVA (three or more levels of X) X and Y are both continuous - CORRELATION (only one X variable at a time - REGRESSION (multiple X variables at a time)
Sample Mean
UNBIASED The sample mean is an unbiased estimator - on average, the sample mean equals the population mean
Response Bias - Item/Option Order
The bias caused by the order in which response items/options are listed To reduce bias: - randomize the sequence of items/options from one respondent to the next (easily done in Qualtrics)
Coding
The process of transforming raw data into numbers Steps in Coding: - specify the categories into which the responses are to be placed - assign code numbers to the classes - keep a record of code values
How to Assess Variation in a Population: Histogram
The shape of the distribution tells us the amount of variations - wider horizontal distribution = more variation Shape: Is the population normally distributed? - Normal distribution is the benchmark for moderate amount of variation in population
What are we measuring in Marketing Research
True Level + Bias = Measured Level
Market Segmentation
We further divide a market in smaller subgroups of consumers (ie market segmentation) To qualify as a segment, a subgroup must: - share one or more common characteristics - have similar product wants/needs - be large enough to be of interest to marketers
Non-Sampling Error - Biases
When the individual participates in the study but provides inaccurate responses - can be due to either instrumentation or response bias
Regression Terminology
X - independent variable (predictor) - influences Y Y - dependent variable (outcome) - is influenced by X b0 - intercept - expected level of Y when X=0 b1 - slope - expected change in Y for a 1-unit change in X R^2 - fit - how well does the regression line fir the dots? e - Error - how much do the dots deviate from the line?
Hypothesis
a guess or prediction that can be tested with QUANTITATIVE DATA
Communication
participants respond to researcher's questions
Levels of Measurement (NOIR)
refer to the format of data being collected each level can be analyzed to different degrees data collected at higher levels CAN be represented at lower levels; however; data collected at lower levels CANNOT be represented at higher levels Lowest to Highest Level of Measurement: Nominal (categorical) Ordinal (categorical) Interval (continuous) Ratio (continuous)
Observation
researcher observes and records data
Marketing Research
•A marketer's formal COMMUNICATION link with the external environment •Process of turning DATA into INFORMATION to ACTION •Purpose is to improve decision-making