MARK 221 Final
Net Promoter Score
"Would you recommend the company (the brand) to a friend?" Results in interval/ratio (scale) data. Scale from 0 to 10 • Promoters: 7-10 • Passives: 0 to 6 NPS = % of Promoters - % of Passives
Marketing Research Function
- Descriptive Function: gather and present statements of fact- Diagnostic Function: explain data/figure our why- Predictive Function: how can firms take advantage of opportunities in the marketplace
Marketing Intelligence
- formal link between the external environment and marketing decision making- intelligence on customer needs and wants, behavioral patterns and impact on business performance
Sample Frame
1. A list of the population elements from which we select units to be sampled (if such a list is available). 2. A specified process
Three different ways to measure a respondent's income:
1. Ask for his or her exact income (e.g., $87,500). 2. Ask him or her to select a range of income (e.g., 1 = less than $25k, 2 = $25k to $50k, 3 = $51k to $100k, 4 = more than $100k). 3. Compare the given respondent to the other respondents and rank them (i.e., assign the respondent with the highest income the number 1, then 2, and so on).
Good Question Wording: questionnaire design
1. Be brief. 2. Focus on a single topic or issue. 3. Be clear. Make sure that the question has a common interpretation. 4. Use conversational language in line with respondents' core vocabulary. 5. Use grammatically simple sentences. 6. Mutually exclusive categories 7. Totally exhaustive categories
To prove causation, the researcher must demonstrate three things
1. Concomitant variation 2. Appropriate time order 3. Elimination of other causal factors
How would you measure brand loyalty? • What dimensions play a role in forming loyalty?
1. Define the Concept Theoretically (Conceptual Definition) • Linked to the problem definition stage of the marketing research process. • Constructs are specific types of concepts that exist at higher levels of abstraction. Some constructs are sometimes called "latent" variables (unobserved) 2. Develop an Operational Definition 3. Develop measurement scale
Developing a Sample Plan
1. Define the relevant population 2. Obtain a sampling frame (for probability sampling plans) 3. Design the plan (needed size, method) 4. Draw the sample 5. Validate the sample 6. Resample, if necessary
Conjoint Steps
1. Design the Conjoint Study 2. Collect Data 3. Select an Estimation Method 4. Evaluate the Results
Criteria for a good questionnaire
1. Does it provide the necessary decision-making information? - It must translate the information needed (MROs) into a set of specific questions that the respondents can and will answer. - The client's approval should be sought (iterative process) 2. Does it consider the respondent? - It needs to fit the language and phrasing to the audience. 3. Does it facilitate data processing? - It should be easy to check the skip pattern logic. - It should be easy to record and code answers.
Bad Question Wording: questionnaire design
1. Don't go beyond the respondent's experience or expertise. • Don't ask for details that cannot be recalled. 2. Don't make assumptions. 3. Don't use ambiguous wording. • Don't use badly designed response categories. 4. Don't ask double-barreled questions. 5. Don't ask leading and loaded questions.
4 Types of Observational Research
1. Natural vs. Contrived Observation • Natural Observation: Observe subjects in their "natural" environment. • Contrived Observation: Observation of an artificial situation in order to speed up data collection, control for extraneous effects, or measure physiological reactions 2. Open vs. Disguised Observation • Open Observation: People know they are being watched. • Disguised Observation: People do not know they're being watched. --> • Observation in the retail store • Observation through the one-way mirror • Social media tracking • In-home use observations 3. Direct vs. Indirect Observation • Direct Observation: Observe current behavior. • Indirect Observation: Infer behavior through some record of past behavior. Also called physical-trace observation. --> • Pantry checks • Garbologist • Wear and tear of tiles in a museum, paths across campus 4. Human vs. Machine Observation Human Observation: A marketing researcher observes people or phenomena. 1. Ethnographic Research • Goal: What makes people do what they do? • Study human behavior in its natural context and conduct interviews. • Fast growing industry in marketing research. 2. Mystery Shoppers • Collect data about customer-employee interactions. • Gather observational data about stores (e.g., cleanliness). 3. One-Way Mirrors • Popular with focus groups and "in-house" research labs (e.g., children playing with toys in a Fisher-Price lab). Machine Observation: A device records aspects of behavior. 1. Traffic Counters • Determines traffic patterns over a particular stretch of road. 2. Nielsen's TV Ratings • Identifies who is watching which TV shows and matches with demographic profiles. • Netflix steaming tracking 3. Physiological Measurement Devices • Measure heart rate, pupil size, brain activity, eye movement, etc. when experiencing some kind of marketing stimuli.
What are the Two Types of Data
1. Secondary Data: Information gathered for another purpose. 2. Primary Data: Information gathered specifically to serve the research objectives at hand.
There are two types of surveys
1. Self-Administered • Paper • Electronic 2. Not Self-Administered • Phone • Personal
Hypothesis Testing
1. State the null and alternative hypotheses. 2. Choose a test and significance level. 3. Compute the observed test statistic. 4. Calculate the p-value. 5. Make a statistical decision (reject or fail to reject the null hypothesis), and interpret the results.
Correlation Testing
1. State the null and alternative hypotheses: H0 : Satisfaction with service and purchase intentions are not positively correlated H1 : Satisfaction with service and purchase intentions are positively correlated (this is a directional hypothesis --> one-sided!)
How would we construct a sampling distribution of the sample mean?
1. Take a sample of size n from the population. 2. Compute the sample average. 3. Plot the sample mean on the x-axis. 4. Repeat.
Reporting: The final report and presentation are important parts of the marketing research project because:
1. They are the tangible products of the research effort. 2. Management decisions are guided by the final report and the presentation (so report findings clearly). 3. The involvement of many marketing managers with the project is limited to the final report and presentation. 4. Management's decision to undertake marketing research in the future or to use the particular research consulting company/group again will be influenced by the perceived usefulness of the final report and the presentation.
Two approaches to marketing research
1. continuous- marketing information system (MIS)-decision support system (DSS)2. focused, one-time projects-marketing research (MR) projects
Univariate Statistics: Single Proportion Examples
A cellular service provider offers 4 different phone plans, each of which have been equally popular across the population of mobile phone users in the past. Is there evidence that this is no longer the case? We could investigate this with a survey question asking customers which plan they prefer. • Question: Are the observed differences in the frequency of each plan due to random sampling, or is there evidence that the population proportions for these three plans are no longer equal? *image shows stat testing
Leading and Loaded Questions: questionnaire design
A leading question includes wording that suggests what the "correct" answer should be. • "How much do you think you would pay for a pair of sunglasses that will protect your eyes from the sun's harmful ultraviolet rays, which are known to cause blindness?" A loaded question is emotionally charged or suggestive of socially desirable answers: • "Do you think that patriotic Americans should buy imported automobiles when that would put American labor out of work?
Questionnaire
A questionnaire is a set of questions designed to generate the data necessary for accomplishing the objectives of the research project. - It provides standardization and uniformity in data gathering. - It allows a valid basis for comparing respondents' answers.
Factor analysis
A statistical procedure that identifies clusters of related items (called factors) on a test; used to identify different dimensions of performance that underlie a person's total score. factor: A linear combination of variables that are correlated with each other The purpose of factor analysis is data simplification. The objective is to summarize the information contained in a large number of metric measures (e.g., rating scales) with a smaller number of summary measures, called factors. As with cluster analysis, there is no dependent variable. • Data simplification (data reduction) technique • Combine some of the variables together in a "factor" • Removes redundancy or duplication from a set of correlated variables • To create a smaller set of relatively independent variables (factors) • Driver analysis: the effort expended to identify those factors that are the root causes of activity costs. Use a representative variable from each factor & Factor score To do the following (examples): • Run regression (identify the most important factors/drivers) • Compare groups on each factor
Sample
A subset of the population of interest.
Confidence Interval
ALL confidence intervals have three components: 1. point estimate (the statistic) 2. critical value of test statistic 3. standard error (SE) • Confidence intervals are more informative than point estimates because they characterize the uncertainty inherent in the estimate. Need to specify: • Size of sampling error (SE) • Confidence/significance level • Expected variance • Sample size • If a = 0.05, then the confidence level is 0.95.
Advantages & Disadvantages of Surveys
Advantages • Ease: Questionnaires are relatively easy to administer to a large number of respondents. • Reliability: Standardization reduces variability in the answers that may be caused by differences in interviewers or the interview process. • Simplicity: Coding, analysis, and interpretation of data. • Surveys have the highest rate of usage when compared to other means of collecting data. Disadvantages • Motivating respondents to respond truthfully and submit the survey. • Structured data collection with fixed-response choices may result in loss of validity for some measures (e.g., beliefs or feelings). • Properly wording questions is not easy. • Getting the right sample.
Advantages and Disadvantages of Stratified Sampling
Advantages • Generally will produce representative samples. • Avoid conscious bias. • We can use statistical formulas to calculate the sample size (i.e., we can estimate sampling error). Disadvantages • Requires a sampling frame, which is not always available. • Often takes longer and costs more to select.
Advantages and Disadvantages of Non-Probability Sampling
Advantages •Samples can be drawn quickly and easily. • No sampling frame is necessary. •Excellent for exploratory research. Disadvantages •Samples can include irrelevant units. • Can't generalize to the population of interest. • Can't evaluate sampling error.
Advantages and Disadvantages of Internet Surveys
Advantages • Cost effective • Fast data collection (instantaneous) • Web design can be sophisticated and flexible (for instance, skip patterns) • Respondent is anonymous • Geographically flexible Disadvantages • Quality of Internet samples varies • Possible self-selection error (voluntary, have to be on website) • Related to the sampling issue
Attitude
An enduring organization of motivational, emotional, perceptual, and cognitive processes with respect to some aspect of our environment. • It is a learned predisposition to respond in a consistently favorable or unfavorable manner toward an object. • It is one of the most pervasive notions in marketing
Bivariate statistical analysis
Analyze two variables simultaneously. The scale level of the two variables (the two questions) determines what statistical technique you use.
Costs of Test Marketing
Approximately $500,000 in direct costs for a simple two market test. Direct Costs • Marketing mix costs (e.g., ads and coupons). • Outside vendors (e.g., marketing research and ad agencies) Indirect Costs • Management time and diversion of attention. • Negative impact on trade and reputation if the test fails. • Negative impact with competitors gaining information.
Reporting: Follow Up
Assisting the Client • The researcher should answer questions that may arise and help the client to implement the findings. • Leverage opportunities to turn follow-up into a new research project. Evaluation of the Research Project • Every marketing research project provides an opportunity for learning and the researcher should critically evaluate the entire project to obtain new insights and knowledge. • Get feedback from the client on the implementation of research findings and the result.
Experimental research is the best form of causal research.
Causal research is designed to determine whether a change in the explanatory or independent variable (IV) likely caused an observed change in the dependent variable (DV).
Univariate Statistics: One-Way Chi-Squared Test
Components of the test: • One categorical/nominal variable (e.g., preferred phone plan); • Several levels (e.g., plan 1, plan 2, plan 3, plan 4)
Why conjoint analysis
Conjoint analysis is a decompositional approach used to analyze consumer preferences. Two main types: • Ratings based conjoint analysis • Discrete choice conjoint analysis Combines multiple features (joint, combined) Forced to make trade-offs • Why do you want to see how people make trade-offs (from the business perspective)? Provides information on what people truly value • Actual choice decisions • Avoids non-conscious overstating of the importance of the attributes • Difficulty of processing of information during trade-off exercise • Mimic (or attempts to) the real-world decision-making process
4 Types of Non-Probability Sampling
Convenience Sampling Snowball Sampling Quota Sampling Judgement Sampling
Correlation
Correlation helps answer questions like: • As X increases, does Y tend to increase or decrease? • If X is greater in value, does Y tend to also be greater in value?
Data entry requirements
Data entry requires:• Data cleaning• Coding• Tabulation prior to statistical tests
Census
Data obtained from every member of the population.
Characteristics of Scales
Description • Unique labels or descriptors that are used to designate each value of the scale (e.g., very good, okay, very bad). Order • Relative sizes or positions of the descriptors - good is better than bad. Distance • Absolute differences between the scale descriptors are known and may be expressed in units. Origin • The origin characteristic means that the scale has a unique or fixed beginning or true zero point.
4. Evaluate the Results
Design and run market simulations using the part-worth estimates. • How would consumers respond to a new product offering? • How would consumers respond to competitor actions? • Estimate market demand and sales. • Profitability analysis if you have data. Directional indicators of market share Preference share, not market share Other possible uses for the part-worth estimates: • Determine attribute importance. Note: The relative importance of an attribute is its share of the total change in utility possible across all attributes. • Segment consumers based upon their part-worth estimates.
Closed-Ended Scaling Techniques
Dichotomous Multiple Choice Rank Order Scale Scaled Response --> Continuous Rating Scale; Itemized Rating Scale
Double-Barreled Questions: questionnaire design
Double-barreled questions address more than one issue: • "Do you think Coca-Cola is a tasty and refreshing soft drink?" • "How satisfied are you with employees being easy to understand and understanding you?"
Examples of Experiments
E-mail subject line: • Up to 60% off + FREE SHIPPING on ALL Outlet! • Shop Up to 60% off + FREE SHIPPING on ALL Outlet! • Sale Up to 60% off + FREE SHIPPING on ALL Outlet! • DVs: Open rate, conversion rate, order size in dollars • The verb "shop" increased order size. Effect sizes: O1 (control) - O2 (Shop) O1 (control) - O3 (Sale) O2 (Shop) - O3 (Sale)
practical considerations for experimentation
Ethical and legal considerations • Facebook emotional contagion experiment: Facebook didn't get consent (was added 6 months later) or ethics board approval • Banks (Redlining) • Loyalty programs; Financial incentives and promotions Internal support • Costs • Timing (other initiatives, resource allocation, etc.) The validity of the results
Univariate Statistics: Single Mean Test
Example: Crest Toothpaste • Crest is introducing a new toothpaste and wants to set the price at $5. They are interested in knowing whether the mean WTP in the population is less than $5 . • Suppose the marketing research team at P&G takes a sample of n = 200 customers and finds that the average WTP in the sample is 𝑋 = 4.3. Further, the sample standard deviation is s = 1.9. • Question: Is the fact that the observed sample mean is less than $5 due to random sampling, or is there evidence that the population mean WTP for Crest's new toothpaste is actually less than $5? *image shows stat testing
Bivariate: Difference in Means
Example: Crest Toothpaste • Suppose the marketing research team at P&G collected information on gender. We could then determine if there is a difference in WTP between males and females. • Question: Is the observed difference in sample means due to random sampling, or is there evidence that the population mean WTP for Crest's new toothpaste is different for males and females? With a p-value less than a= 0.05, we can reject the null hypothesis. Hence, there is evidence to conclude that the average WTP for Crest's new toothpaste is different for males and females. *image shows stat testing
Bivariate: Two-Way Chi-Square Test
Example: General Mills Cereal • General Mills is thinking about introducing an organic option into its current line of cereals. The CMO wants to know if there is a relationship between age and the intent to purchase organic cereal. • The marketing research team takes a survey of n = 500 respondents between the ages of 20-70 and asks whether or not they intend to purchase General Mills' new organic cereal. *image shows stat testing
Hypothesis testing for regression model
Example: Regression: A record company boss was interested in predicting record sales from advertising • Sample size (n) = 200 different album releases • Outcome variable (Y) = Sales (CDs and Downloads) in the week after release • Predictor variable (X) = The amount (in £s) spent promoting the record before release *image shows stat testing
3 Major Research Designs
Exploratory: If little is known about the problem. Descriptive: If the problem is "somewhat" clear. Causal: If the problem is very clear
Know how to read and interpret output tables from SPSS
F = observed test statistic Sig = p-value
Link Between Attitudes and Behavior
Favorable attitudes tend to lead to higher usage and less favorable attitudes tend to lead to lower usage. • Unfavorable attitudes tend to lead to stopping usage. • Attitudes based on actual experience are more highly correlated with behavior than attitudes based on advertising exposure.
Central Limit Theorem
For any population, regardless of its distribution, any associated sampling distribution approaches a normal distribution as the sample size increases.
For nominal/ordinal data use...
Frequency table Relationship between two nominal/ordinal variables: Cross tabulation (contingency table)
Descriptive statistics: Summarizing Data
Graphical summaries • Histogram • Pie chart, bar chart, line graph, etc. • Scatterplots Numerical summaries • Measures of central tendency (mean, median, mode) • Measures of dispersion (standard deviation, variance, range) • Relational measures (correlation, covariance)
Descriptive statistics: Measures of central tendency: Sample Median
If n is odd then Median = Xj where j= n+1/2 •If n is even then Median m is m= (Xj + Xj+1)/2 where j=n/2
Bivariate: ANOVA Test
If there are more than two groups, we can use ANalysis Of VAriance (ANOVA). To use an ANOVA model, we must have • one interval/ratio dependent variable • one or more nominal/ordinal independent variables (sometimes called factors). Examples: • Do various segments differ in terms of their volume of product consumption? • Do the brand evaluations of groups exposed to different commercials vary? • What is the effect of consumers' familiarity with the store (measured as high, medium, and low) on preference for the store? Example: Movie Data • Using the movie data from lab 1, we can explore the differences in movie attendance (Q3) by class rank (Q13). • We have K = 5 groups (freshman, sophomore, junior, senior, grad). *image shows stat testing
Observational Research & 4 Types of Observational Research
Instead of asking respondents directly (surveys) researches observe how respondents behave. Methods can be classified along four dimensions: 1. Natural vs. Contrived Observation 2. Open vs. Disguised Observation 3. Direct vs. Indirect Observation 4. Human vs. Machine Observation
3. Select an Estimation Method
It depends on the type of conjoint being conducted. • Regression is most common for ratings-based conjoint. • Hierarchical Bayesian methods are commonly used in choice-based conjoint.
Writing Reports
Know your audience • A report should be written for a specific reader or readers (e.g., the marketing managers who will use the results). • The appendix is designed to provide technical details Make it easy to follow • The story and logic in the report should be easy to follow, with both clear writing and a clear structure. Be professional • How a report looks is important, a signal of quality and competency. Objectivity • Be honest and up-front. Don't oversell the results, be aware, but not afraid of risks Reinforce text with visuals • It is important to reinforce key information in the text with tables, graphs, pictures, maps, and other visuals. Be concise • A report should be concise, but brevity should not be achieved at the expense of completeness (i.e., use the appendices liberally). Don't report statistical hypotheses (the statements)! Put the statistical output in appendices • Inserting output tables from SPSS in the body of the report does not add any value. Use the appendices to include detailed information on the statistical procedures. Report: 1. observed test statistic, 2. p-value 3. each with an accuracy of two decimal places
Types of Experimental Research
Laboratory Experiments: Experiments conducted in a controlled, laboratory setting. • High or low internal validity? • High or low external validity? Field Experiments: Tests conducted outside the laboratory in an actual environment, such as a marketplace. • High or low internal validity? • High or low external validity?
comparison of marketing information system (MIS) vs marketing research (MR)
MIS:- continuous and ongoing- future oriented- systematic- internal and external information MR:- ad hoc, piecemeal, each project has a beginning and end- current decision based on pertinent information- flexible and situation specific- mostly external
Usage of conjoint analysis examples
Marketing • New Product Introduction • Pricing Research • Brand Equity Research • Market Segmentation • Product Positioning/Line Extensions Other applications • Employee Research (benefits, retention, etc.) • Litigation (assessment of damages) • Environmental impact studies • Patient/physician communications • Transportation research
Descriptive statistics: Numerical measures: Relational Measures
Measuring the relationship between two interval/ratio variables: • Graphical Summary: Scatterplot • Numerical Summary: Sample correlation/covariance
For categorical data use...
Numerical summaries: • Frequency tables • Counts • Sample proportions
social desirability responding
People like to appear socially desirable. The presence of an interviewer influences answers • Bias in responses caused by respondents' desire, either conscious or unconscious, to gain prestige or appear in a different social role, even when an interviewer isn't present. • "Porsche is considering adding another hybrid vehicle to its current line of models. How interested would you be in such an option?"
First Step: Marketing Research Process Planning:1. Problem Identification Situation Analysis --> the "why" --> exploratory research methods
Pilot studies • conducting mini studies that may reveal problem areas Experience survey (key informant survey) • discussions with knowledgeable persons both inside and outside the firm Case analysis • looking at examples of former but similar situations• Focus groups • having small groups discuss topics such as the company's product or service Secondary data analysis
Marketing Research Process
Planning 1. problem identification 2. determine the research design Research Implementation 3. design the data collection method 4. sampling and data collection Developing Recommendations 5. analyse and interpret the data 6. prepare and report the findings
Projection Technique Example: Sentence Completion
Please complete the sentences as quickly and spontaneously as possible:
Sampling Methods: Probability vs Non-Probability Sampling
Probability Sampling • Objective • The probability of selection of an element is known and non-zero. • Can calculate sampling error. • Strict procedures to follow. • Yields a representative sample Non-probability Sampling • Subjective procedure. • Probability of selection of an element is unknown • Cannot calculate sampling error • The sample not always representative.
You can collect data on importance ratings of different attributes
Problems: • Single attribute at a time • Low discrimination • Not very actionable
Measurement Error
Processing/Input errors Instrument errors Interviewer errors Response errors • Non-response bias • Careless responding • Response bias 1. Social desirability responding 2. Acquiescence 3. Extreme response style 4. Midpoint responding
Regression SPSS
R Square = R^2 Sig = P-value
Ratio Scales
Ratio scales possess all of the properties of the nominal, ordinal, and interval scales, plus a natural zero (i.e., you don't possess anything of the measured attribute if you get a zero on this scale). What can we calculate with a ratio scale? • All mathematical operations (e.g., mode, median, mean). Examples: • How old are you? _____ • How many people are in your class? _____ • How many miles do you travel to school?_____
Scaled Response Questions
Respondents choose a category to describe the intensity of their response. Basically the same as multiple-choice, except: • Choices cover a range of intensities. • Results in interval or ratio data. • Continuous rating scale questions have respondents manipulate some kind of graphical input. Common types of itemized rating scale questions: • Likert • Semantic differential • Purchase intent
Continuous Rating Scale Questions
Respondents rate the objects by placing a mark at the appropriate position on a line that runs from one extreme of the criterion variable to the other. (slide ranking between 0 and 100 etc)
Response Rate
Response rate = # 𝑟𝑒𝑠𝑝𝑜𝑛𝑑𝑒𝑛𝑡𝑠 / 𝑆𝑎𝑚𝑝𝑙𝑒 𝑠𝑖𝑧e
Descriptive statistics: Measures of dispersion: Sample Variance
S^2
2. Collect Data
Sample respondents from your target market (population of interest). • Pre-test your study and make necessary adjustments to attributes and levels before launching it to your entire sample. Design a data collection procedure. • Type of conjoint (ratings vs. choice based) • Pictorial vs. verbal descriptions of products
1. Design the Conjoint Study
Select attributes relevant to the product or service category as well as the levels for each attribute. • Attributes: Number of attributes, Size, Brand, Performance • Levels for each attribute: Categorical measures of attributes. Define how many levels for each: Size (Small, Medium, Large), Brand (A,B,C), Performance (Fast, Medium, Slow) • Exploratory research (e.g., focus groups) and text analysis might be especially helpful. Develop the product bundles (profiles, concepts) to be evaluated. • Respondents will not be able to evaluate each possible profile; for example, if we have 5 attributes with 3 levels each, there are 35=243 possible profiles (combinations of all attribute levels) • Instead, we typically employ a fractional-factorial design ( a type of experimental design) • Guideline: No more than 30 profiles should be evaluated by each respondent.
Probability Sampling: Simple Random vs Systematic Sampling
Simple Random Sampling - assign a number to each element - randomly select unit Systematic Sampling - Assign numbers and produce a skip interval "k" - Choose every kth element in the sampling frame
4 Types of Probability Sampling
Simple Random Sampling Systematic Sampling Stratified Sampling Cluster Sampling
Multivariate (MV) analysis
Simultaneous use of multiple measurements on each individual or object being studied
MRO First Step: Marketing Research Process Planning: 1. Problem Identification
Statements of precise information needed to address marketing decision problem Drive selection of research tools to address marketing decision problems • Precision is important
Discriminant Analysis
Statistical methods that use one or more predictor variable(s) to discriminate between categories (dependent variable is categorical).
Discriminant analysis (DA)
Statistical methods that use one or more predictor variable(s) to discriminate between categories (dependent variable is categorical). ▪Determine if there are statistically significant differences between the average discriminant score profiles of two (or more) groups (in this case, users and nonusers). ▪Establish a model for classifying individuals or objects into groups on the basis of their values on the independent variables. ▪Determine how much of the difference in the average score profiles of the groups is accounted for by each independent variable. Regression: • Dependent variable is metric (ratio or interval) • Independent variables: any • Ex: the likelihood of purchase as DV DA: • Dependent variable is non-metric (nominal, categorical) • Independent variables: metric only • Ex: buyer versus non-buyer DV Idea: • Find a set of independent variables (factors) to predict the outcome. • In DA, predict the membership to the category or discriminate best among two or more groups.
Probability Sampling: Stratified vs Cluster Sampling
Stratified Sampling (proportional or disproportional) • Split the sampling frame into mutually exclusive groups • Simple random or systematic sampling on the subgroups separately Cluster Sampling • Split the sampling frame into mutually exclusive, relevant subgroups • Select a random sample of the subgroups Ex: Choose from 366 Metropolitan statistical areas (MSA)
Questions Survey Research Helps Answer
Survey research is popular because it can be used to answer what, who, and how questions. • What stores are frequented? • What is the demographic and lifestyle profile of our customers? • Who are the users of our category? • How do people make a purchase decision?
Reporting: Talking about future research
Talk about some limitation of the study • Briefly about the reasons • Show enthusiasm to do it differently the next time • This leads you to future research opportunities Ideas for future research • Use what you saw that was outside of the scope of the project • Link it current MDPs • Link to other possible MDPs
Test Markets
Test marketing refers to field testing a new product or other changes to the marketing strategy. • Use a single market, group of markets or a region of the country. • Involve the use of experimental procedures. Usage & Objectives • Estimate market share and volume. • Estimate effect on other items that the company markets (i.e., the cannibalization rate). • Collect data about potential customers. • Analyze the reactions and behavior of competitors. Four Factors to Consider: 1. What is the tradeoff of the costs and risks of product failure and the potential profits and probability of success? 2. How quickly can competitors respond to or copy the product? 3. Consider the investment required to produce the product for the test market. 4. Appraise the impact of product failure on the company's reputation.
sample
The ____ is a subset of the population. • Hopefully representative • Collecting a sample gives us information (data) about the population
Z-test vs. t-test
The choice depends on: • whether the population standard deviation is known; • the sample size: t for small (<30 observations) z for large samples
Factor score
The composite measure created for each observation on each factor extracted in the factor analysis.
parameters
The goal in marketing research is to learn about certain features of the population, which are called ____
Inferential Statistics: Point Estimation
The goal of point estimation is to find a single number (i.e., a statistic) to best represent the parameter of interest. Statistical inference is about using data to systematically learn about population parameters. Problem: We are representing our "answer" with a single point. • There is uncertainty in any statistic based on a random sample. • Rather than report a single number, perhaps we should report a range of values that we think are "plausible." • This is the idea of a confidence interval
Descriptive statistics: Measures of central tendency: Which measure of central tendency is the best to use?
The median is more robust (i.e., less affected by outliers) than the mean. Therefore, we should really only use the sample mean as a measure of central tendency when the distribution of our sample is symmetric.
Incidence Rate
The number of respondents from a sample pool that will qualify for your study. If your market research survey is looking to target females only, your incidence rate will immediately drop from 100% to ~50% Incidence rate = 𝑇otal 𝑀arket / 𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜n of Interest
Sampling distribution
The probability distribution of a statistic is called the ________. It is a theoretical distribution, as it represents a distribution of all other things that could happen.
p-value
The probability of seeing something as or more extreme as what we observed, given that the null is true. We can now answer the question: How extreme is our sample, assuming that the null is true? • Smaller p-values imply more evidence against H0.
First Step: Marketing Research Process Planning: 2. Determine the Research Design
The research design specifies methods and procedures for collecting and analyzing the needed information.
Least squares estimation
The technique used to fit data for X and Y to a line that best represents the relationship between the two variables The best line has the smallest Sum of Squared Errors (SSE)
Population
The total group of people from whom information is needed. (universe, population of interest)
Scaling techniques
These techniques cover the different ways to go about measuring the concepts of interest. The techniques fit under two broad categories: open and closed-ended. Open-Ended • Allows for unaided recall. • Respondents reply in their own words (qualitative) or with numeric input (ratio data). Closed-Ended • Structured with aided recall. • Respondents choose from a list of answers (quantitative).
statistic
This information from a sample can be summarized through a ______. • Estimation of a population parameter • Goal: make inferences about parameters
Reporting: Using Tables
Title and number • Every table should have a number and a title. Arrangement of data items • The arrangement of data items in a table should emphasize and/or point to the most significant aspect of the data. Basis of measurement • The basis or unit of measurement should be clear. Good design • Make the table easy to read (i.e., follow good design principles). Footnotes • Information that cannot be incorporated in the table should be explained by footnotes. Sources of the data • If the data contained in the table are secondary, the source of data should be cited.
Survey Error: Total Error Components
Total Error > -Random Error -Systematic Error- sample design error, measurement error
Exploratory Research First Step: Marketing Research Process Planning:2. Determine the Research Design
Used in Focus Groups/Interviews Exploratory research is the term used to refer to preliminary research used to clarify the exact nature of the problem to be solved. Tends to be flexible, fast, and relatively cheap (i.e., small samples). Uses unstructured data collection, resulting in qualitative data. It helps guide future research (i.e., does not provide conclusive results).
Casual Research First Step: Marketing Research Process Planning: 2. Determine the Research Design
Used in lab/field experiments It provides the answer to why. Uses large, representative samples. Uses structured data collection, resulting in quantitative data. It provides conclusive results that inform decision making.
Inferential statistics
Used to draw conclusions and make inferences about the population from the sample. • Confidence intervals, hypothesis testing, etc
Descriptive (summary) statistics
Used to summarize information in a sample. • Reporting (observed) sample mean, variance, correlation, etc.
Validity
Validity is measuring what you intended to measure. Two types of experimental validity: • Internal Validity: The extent to which competing explanations for the experimental results observed can be avoided. "Do you measure what you want to measure?" • External Validity: Whether the causal relationships measured in an experiment can be generalized to outside persons, settings, and times. • "Can you extend the results from your experiment to the real world?" To get a valid experiment, control for extraneous variables.
Multiple Regression: Assumptions
Variable Type: • Outcome should be continuous • Predictors can be continuous or dichotomous Non-Zero Variance: • Predictors must not have zero variance Linearity: • The relationship we model is, in reality, linear Independence: • All values of the outcome should come from a different person Homoscedasticity: • For each value of the predictors the variance of the error term should be constant. Independent Errors: • For any pair of observations, the error terms should be uncorrelated. Normally-distributed Errors No Multicollinearity: • Predictors must not be highly correlated.
Descriptive statistics: Measures of dispersion
We want to know the variation/spread of values around the mean. Three measures of dispersion: • Sample variance • Sample standard deviation • Sample range
First Step: Marketing Research Process Planning: 1. Problem Identification
Why is the info being sought? Iceberg Principle: (the why): observe symptoms and use them to get to the root of the "why"/problem situation analysis (exploratory research): pilot studies, experience survey, case analysis, focus groups, secondary data analysis
semantic differential scale
a five-point scale in which the opposite ends have one- or two-word adjectives that have opposite meanings The negative adjective typically appears on the left side of the scale. • This controls the tendency of some respondents to straight-line
Standard Normal Distribution
a normal distribution with a mean of zero and a standard deviation of one
Likert Scale
a numerical scale used to assess attitudes; includes a set of possible answers with labeled anchors on each extreme consist of a series of statements that express either a favorable or unfavorable attitude toward a concept. • Example: Going to amusement parks is one of my favorite activities. ( ) Strongly agree (5) ( ) Agree (4) ( ) Neither agree or disagree (3) ( ) Disagree (2) ( ) Strongly disagree (1) • Results in interval data.
An operational definition
a precise definition that specifies the activities or operations to measure the concept. • The concept is broken down into • measurable components • method for creating an overall value based on those components
Purchase Intent Scale
a scale used to measure a respondent's intention to buy or not buy a product --> some of the most commonly used scales in marketing research. When combined with estimates about actual purchase likelihood, they can be the basis for go-no-go decisions for a product launch. • Example: If this coffee maker sold for approximately $75 and were available in the stores where you normally shop, how likely would you be to buy it? ( ) definitely will ( ) probably will ( ) might or might not ( ) probably will not ( ) definitely will not
Type 2 error
accepting the null hypothesis when it is false --> false negative
Summary of Bivariate Statistics
chart shows what bivariate methods work for each kind of data
Summary of Univariate Statistics
chart shows what descriptive and inferential methods work for each kind of data
What to look at before doing marketing research
cost/benefit analysis, timing, control of decision, manager's point of view
Segmentation Analysis
dividing up customers into groups or segments based on similar needs or wants
MDPs First Step: Marketing Research Process Planning: 1. Problem Identification
linked to the tactics to support a managerial or strategic marketing decision problem. Often related to the 4 P's and are congruent with marketing strategies
Strategic Use of Marketing Resources
marketing strategy guides the long-term use of the firm's resources based on the firm's existing and potential internal capabilities and on projected changes in the external environment
Four Types of Measurement Scales
nominal, ordinal, interval, ratio
For a hypothesis test of level alpha, the decision rules take the following form:
reject H0 if p-value is <= alpha fail to reject H0 if p-value > alpha
Type 1 error
rejecting the null hypothesis when it is true; finding a statistically significant effect when no true effect exists --> false positive *False positives (Type I errors) are often thought to be "worse" (more costly) than false negatives (Type II errors).
Measurement
the process of assigning numbers or labels to objects, persons, states, or events in accordance with specific rules to represent quantities or qualities of attributes.
Perceptual Maps
tool used to depict graphically the positioning of competing products Multidimensional scaling (MDS) addresses the problem of positioning objects in a perceptual space • Determination of dimensions • Meaning of the dimensions • Position of objects (brands) with respect to these dimensions • Preference of the consumers (ideal points) Input: similarity judgments on different attributes Outcome: perceptual map Perceptual Map Use: • Strategic competitive analysis • Product life-cycle analysis • Market segmentation • Evaluation of advertising • Store image research • Brand-switching research • Attitudes research
Reporting: Bar Chart vs Histogram
• A bar chart displays data in various bars that may be positioned horizontally or vertically. • The histogram is a vertical bar chart in which the height of the bars represents the relative or cumulative frequency of occurrence of a specific variable.
Reporting: Line Chart
• A line chart connects a series of data points using continuous lines. • This is an attractive way of illustrating trends and changes over time. • Several series can be compared on the same chart, and forecasts, interpolations, and extrapolations can be shown.
Causation => Regression Analysis
• A way of predicting the value of one variable from another • It is a hypothetical model of the relationship between two variables • The model used is a linear one • Therefore, we describe the relationship using the equation of a straight line
Concrete concepts can often be measured with a single question (unidimensional). Examples include:
• Age • Gender • Income
Running Multiple Regression
• All variables are entered into the model simultaneously • The results obtained depend on the variables entered into the model • It is important, therefore, to have good theoretical reasons for including a particular variable Generalization • When we run a regression, we hope to be able to generalize the sample model to the entire population. • To do this, several assumptions must be met • Violating these assumptions stops us from generalizing conclusions to our target population.
Experimental Design
• An experimental design is a set of procedures that define the experiment. • It includes four factors: 1. The IV, also known as the treatment variable. 2. The subjects assigned to "conditions" or "treatments" 3. The DV (what is being measured). 4. The plan or procedure to deal with extraneous variables.
In-depth Interview
• An unstructured, direct, personal interview in which a single respondent (i.e., one-on-one) is probed by a highly skilled interviewer to uncover underlying motivations, beliefs, attitudes, and feelings on a topic Major Advantage: Can uncover deep insights about underlying motives, in particular when topic is complex (probing questions). Major Disadvantages: • The skilled interviewer is essential - they may be hard to get and they influence the discussion. • The technique results in very rich qualitative data that can be hard to interpret. • A very small sample is used so you can't generalize findings.
Measuring Attitudes
• Attitudes are not directly observable, but are measurable by indirect means such as verbal expression or behavior. • We can measure attitude along three dimensions: 1. Affective: The feelings or emotions toward an object. • Example: "Seeing a BMW makes me happy." 2. Cognitive: Awareness or knowledge about an object. • Example: "My BMW gets great gas mileage." 3. Behavioral: Reflects intentions, a predisposition to act. • Example: "I will always buy a BMW ."
• Multiple questions to measure constructs (multidimensional). Examples include:
• Brand loyalty • Satisfaction • Attitude --> Evaluation, strength, cognition/affect, etc.
Why do regression?
• Determine whether the independent variables explain a significant variation in the dependent variable • Determine whether a relationship exists between the two variables (as with correlation) • Determine how much of the variation in the dependent variable can be explained by the independent variables • Determine the strength of the relationship (also addressed with correlation) • Determine the structure or form of the relationship • Determine the mathematical equation relating the independent and dependent variables • Predict the values of the dependent variable • Control for other independent variables when evaluating the contributions of a specific variable or set of variables • Controlling for extraneous variables starts to sound like we're able to say something about causation, but...Regression analysis is concerned with the nature and degree of association between variables and does not imply or assume causality. It is assumed by the researcher.
Focus Groups (Qualitative Research)
• Exploratory Research • A group of participants who are led by a moderator in an in-depth discussion on one particular topic or concept. • Unlike in-depth interviews, they allow for group dynamics What they do: • generation of ideas/attitudes/perceptions • testing vocab • reveal consumer needs/motivations/perceptions and attitudes about products/services • tests survey instrument items • clarify the findings of quantitative studies Advantages: • They are easy to execute and very flexible. • They produce rich information due to group and moderator interaction. • They provide a good platform for idea generation, brainstorming, and understanding customer vocabulary. Disadvantages: • They are not representative and thus hard to generalize. • The data is hard to analyze • The moderator makes or breaks it.
Cluster Analysis
• Exploratory analysis. "Unsupervised" analysis • Classify objects (respondents) or people into groups according to their similarities on some criteria • Data: group rows, not columns (as in factor analysis) • Most used in segmentation analysis • Subjective analysis • Number and subsets of variables impact results • Use: Compare cluster profiles to identify differences ▪Cluster 1 includes those people who do not frequently eat out or frequently eat at fast-food restaurants. ▪Cluster 2 includes consumers who frequently eat out but seldom eat at fast-food restaurants. ▪Cluster 3 includes people who frequently eat out and also frequently eat at fast-food restaurants.
Relationships with Variables and Types of Research
• Exploratory research: Which variables to use • Descriptive: Associations among variables • Causal: Which variables impact which variables
Nominal scales
• Give a unique name or label to a category. • Does NOT show rank order. • Does NOT have equal intervals. What can we calculate with nominal data? • Percentages • Mode: Most frequently occurring response. Examples: • What is your gender? 1. Male 2. Female • Where did you last buy toothpaste? 1. In a supermarket 2. In a discount store 3. Somewhere else
Reporting: Pie Charts
• In a pie chart, the area of each section, as a percentage of the total area of the circle, reflects the percentage associated with the value of a specific variable. As a general guideline, a pie chart: • Should not require more than seven sections; • Should move clockwise from the largest to the smallest section.
Methods to Improve Response Rate
• Incentives --> monetary • presents (e.g., pens, pencils, stickers) • Personalization --> personalized letter • send a follow-up postcard • phone call • email • Make it easy --> email with a hyperlink • self-addressed, stamped envelope
We can measure careless responding using:
• Instructional manipulation checks (IMC) • Self-report measures of response quality • Response times • Post-hoc response consistency indices (too much or too little)
Presence/absence of an interviewer
• Interviewer Pros: To help/motivate respondent. • Interviewer Cons: Social desirability errors, anonymity is gone, only small samples, geographically constrained, more expensive.
Examples of research questions:
• Is the true average WTP different from the established price of $60 ( <> 60)? • Does the true proportion of satisfied customers exceed 50% (p > 0.50)? • Does advertising have a positive effect on sales (𝛽𝑎𝑑 > 0)? We specify two competing (and mutually exclusive) hypotheses: • Null hypothesis (Status quo): H0 • Alternative hypothesis (research hypothesis): Ha or H1 *Note: These are NOT statements about the data! Correct hypotheses H0 : u = 60 H1 : u ≠ 60 • If a = 0.05, then the confidence level is 0.95.
Selecting a Test Market
• List of markets • Avoid unusual demographics • Consider regional differences. • Sales of that type of product should be typical. • Little media spillover between markets. • Moderately sized markets. • Generalizable distribution channels and competitive situation. • Differing cities should have similar demographics.
Projection Technique Example: Third Person Technique (3rd person technique)
• Mercedes-Benz wants to know how people view the Mercedes brand. Two candidate questions for potential buyers: 1. Why would you buy a Mercedes? 2. Why do you think your friend would buy a Mercedes?
Interval Scales
• Numbers from an interval scale not only represent a classification and an order but also measure distance in units of equal intervals. What can we calculate with an interval scale? • Percentages. • Mode: Most frequently occurring response. • Median: 50th percentile response. •Mean: Average of the distribution. Example: How likely are you to buy a new VW this year? 1. Very likely 2. Likely 3. Neither likely or unlikely 4. Unlikely 5. Very unlikely
Ordinal Scales
• Numbers from an ordinal scale indicate the relative positions of the objects. In other words, ordinal scales preserve order. • Measures go from highest to lowest, most to least, strongest to weakest, etc. • Does NOT indicate the magnitude of differences between objects Does NOT have equal intervals. What can we calculate with ordinal data? • Percentages. • Mode: Most frequently occurring response. • Median: 50th percentile response. Example: • How many movies did you see last month? 1. None 2. One or two 3. Three or more
Descriptive statistics: Measures of central tendency: Sample Mean
• Oftentimes we want to understand the value that is in the "middle" or is "most typical" in our sample. • Sample Mean X = random variable n = total number of observations (sample size) • Scale from 1 to 5. Collected 82 responses. Frequency of responses: 1: 10 2: 15 3: 25 4: 20 5: 12 (1*10+2*15+3*25+4*20+5*12) / 82 = 3.1097561
Non-Probability Sampling: Snowball (referral) Sampling
• Operational Plan: Find one person who fits the characteristics of interest, then ask that person to generate names of others with the same characteristics. • Representativeness: Unlikely to be representative.
Non-Probability Sampling: Judgement Sampling
• Operational Plan: Find willing participants based on researcher's judgment criteria. • Representativeness: Can be representative if researcher uses high quality criteria and adheres to it strictly
Non-Probability Sampling: Convenience Sampling
• Operational Plan: Find willing participants based on the researcher's convenience. • Representativeness: Can be representative if chosen carefully, but can also be completely unrepresentative.
Non-Probability Sampling: Quota Sampling
• Operational Plan: Find willing participants to represent various characteristics known about the target population. For example, 50/50 male/female. • Representativeness: Can be representative if researcher uses high quality criteria and adheres to it strictly.
Projection Technique Example: Word Association
• Please respond to the following words with the first word that comes to mind: • Mercedes-Benz • Tide • Pillsbury
Some housekeeping:
• Population distribution: Frequency of all elements of the population. • Sample distribution: Frequency of all elements within an individual sample (drawn from a particular population). • Sampling distribution (distribution of means): Probability distribution for the set of all possible sample means (of a given sample size, drawn from a particular population).
Sample Design Error
• Population mis-specification • Incorrect sampling frame • Improper selection • Respondents are not in sampling frame • Screening mechanism isn't working properly • Could be due to field worker error
True Experimental Designs
• Pretest-posttest control group design (before-and after with control group): Experimental Group: $55.6 X $75.6 Control Group: $54.2 $62.2 Treatment effect: (75.6-55.6) - (62.2-54.2) = $12 • Posttest-only control group design (after only with control group): Experimental Group: X $75.6 Control Group: $62.2 Treatment effect: (75.6-62.2) = $13.4
Testing the Regression Model: R2
• R2 (Coefficient of Determination) • The percentage (so a value from 0 to 1) of the total variation in Y explained by X • The Pearson Correlation Coefficient Squared in simple R^2= SSR/SST • SSR is Regression Sum of Squares (explained) • SST is Total Sum of Squares (total)
Rank Order Scale Questions
• Rank order scale questions ask respondents to rank a list of alternatives. Example: Please rank the following plush toys with 1 being the brand that best meets the characteristics being evaluated and 3 being the worst brand on that characteristic. appearance huggability size Mattel ( ) ( ) ( ) Fisher-Price ( ) ( ) ( ) Calico ( ) ( ) ( ) • Results in ordinal data.
Measurement Accuracy
• Reliability: Degree to which measures are consistent. • Validity: Degree to which we measure what we were trying to measure.
Protective Techniques
• Respondents are asked to interpret the behavior of others • In interpreting the behavior of others, respondents indirectly project their own motivations, beliefs, attitudes or feelings into the situation and are more likely to speak the truth. • Useful in situations where respondents are unwilling or unable to answer questions truthfully.
Dichotomous Questions
• Respondents choose between two answers. Example: • Circle your gender: (1) Male (2) Female • Usual answers: • Example: Yes/No, Agree/Disagree, Greater Than/Less Than • Results in nominal data. Prone to a large amount of measurement error because alternatives are polarized and a wide range is missing.
Problems with Careless Responding
• Respondents do not always read instructions, wording of the question, and/or the response options. • Respondents may form expectations about what is being measured and respond to individual items based on their overall position concerning the focal issue (i.e., haloing), rather than specific item content. • This can result in inconsistent responding to reverse-keyed items (when you swap the label positions of 'very good' and 'very bad').
Multiple Choice Questions
• Respondents select from a list of more than two answers. • Can select one answer or check as many as apply (also called "Pick Any/J"). Example: In the last three months, have you used Noxzema... (please circle all that apply) • as a facial wash 1 • for moisturizing skin 2 • for treating blemishes 3 • for cleansing skin 4 • for treating dry skin 5 • Results in nominal data.
Measurement Scales
• Scaling refers to techniques used to determine quantitative measures of subjective or abstract concepts (e.g., constructs). • It involves creating a scale (i.e., continuum) used to measure attributes.
Funnel Approach to Layout: questionnaire design
• Screen to get at the right sample of respondents. • Ask easy/general questions to build rapport. • Ask tough/important questions once the respondent has committed. • Ask sensitive and demographic questions at the end to avoid making the respondent uneasy early.
A theoretical definition of a concept
• States the central idea • Defines the concept in terms of other concepts • Differentiates from other concepts • Establishes boundaries (sandbox)
Aacquiesce
• Tendency to agree with items regardless of content. • Also called "yea-saying" or positivity. • Wording is important - avoid vagueness or words encouraging yea-saying. • The opposite also exists, "disacquiescence bias": • Tendency to disagree with items regardless of content. • Also called "nay-saying" or negativity.
Extreme Response Style
• Tendency to endorse the most extreme response category regardless of content. • Manifestation of rigidity, intolerance of ambiguity, dogmatism, and might be related to anxiety and cognitive development. • May result from misfit of scale on survey (for example a 5-point scale) to the subjective inner scale (subject thinks in terms of a 12-point scale). • Does the daycare communicates with parents about day activities? Yes/No
Midpoint Responding
• Tendency to use the middle or neutral category in a rating scale regardless of item content. • If the item content might encourage midpoint responding, consider not allowing for a midpoint response.
Extraneous Variables
• Testing Effect: An effect that is a byproduct of the research process itself. • Instrument Variation: Changes in measurement instruments (e.g., interviews or observers) that might affect measurements. • Mortality: Loss of test units or subjects during the course of an experiment which might result in non-representativeness. • History: Intervention, between the beginning and end of an experiment, of outside variables that might change the dependent variable • Selection Bias: Systematic differences between the test group and the control group due to a biased selection process (i.e., not randomized). • Maturation: Changes in subjects occurring during the experiment that are not related to the experiment but which might affect subjects' response to the treatment variable. • Regression to the Mean: Tendency of subjects with extreme behavior to move toward the average for that behavior during the course of the experiment.
1. Concomitant Variation
• The IV and the DV must vary together in some predicable fashion. This could be a positive relationship • Ex., an increase in disposable income together with an increase sales of luxury cars. This could be a negative relationship • Ex., an increase in disposable income together with a decrease in sales of low quality hamburger.
Tools used in exploratory, descriptive, and casual Second Step: Marketing Research Process Research Implementation 3. design the data collection method
• The basic tool used in exploratory research is the interview guide or observation form. • The basic tool used in descriptive and causal research is the questionnaire (or observed choices and decisions).
2. Appropriate Time Order
• The change in the IV must precede the change in the DV. • A cannot cause B if A does not occur before B does. A then B (necessary condition, but not sufficient) • Ex., snow causes people to miss class
How good is the Regression Model?
• The regression line is only a model of reality, this model is not reality • The questions is: Does the model do a good enough of a job reflecting reality? • We need some way of testing how well the model fits the observed data (reality) • How would we test this model? --> R^2
3. Elimination of Other Causal Factors
• The researcher should be able to eliminate any other potential explanations to account for the change in the DV. This is the most difficult thing to demonstrate. • A study is considered confounded if there is more than one IV that could have caused the effect. There are four basic approaches to controlling extraneous causal factors: 1. Randomization in assigning subjects to treatment conditions. 2. Physical control of the extraneous factor (i.e., holding it constant). 3. Design control of extraneous factors through the specific type of experimental design used. 4. Statistical control through identifying and measuring the effects of the extraneous factors throughout the experiment. Examples • Internet sale in February soared compared to previous years. We must have done the marketing campaigns correctly. • The weather was significantly worse this winter compared to the weather previous years, so people could not drive to the store, so they shopped online. Examples • Google 2000. First-ever tested the number of results the search engine displayed per page, which then and now defaulted to 10, was optimal for users. To 0.1 percent of the search engine's traffic, they presented 20 results per page; another 0.1 percent saw 25 results, and another, 30. Experimental results were bad for longer pages. • Technical glitches that made experimental pages load much longer than the control (10 results) pages.
3 types of statistical Tests
• Univariate: Examine a single variable. • Bivariate: Examine the relationship between two variables. • Multivariate: Examine the relationship between more than two variables.
Random Error
• What is random error? • Differences in the results between the population of interest and the sample. • Also called "random sampling error." • Unless a census is done, there is always a certain amount of random error, but it can be estimated.
Systematic Error
• What is systematic error? • It results from mistakes or problems in the research or sampling design. • Also called "non-sampling error." • Unlike random error, systematic error must be dealt with.
(Pearson's) correlation coefficient
• When you want to describe the relationship between two interval/ratio variables X and Y, use the (Pearson's) correlation coefficient. • The correlation coefficient will always lie between -1 and 1 • Correlation is a measure of a linear relationship between two variables • R^2 is the square of the correlation coefficient and represents the percent of the variability in Y explained by X.
Experimental Design Notation
• X indicates the exposure of a subject to a change in the IV. • O indicates the DV is measured (Observation). • Different time periods are represented horizontally and are indicated with subscripts. • Different subject groups measured at the same time are represented vertically. • Pretest-posttest control group design (before-and-after with the control group): Experimental Group: R O1 X O2 Control Group: R O3 O4 Treatment effect: (O2 - O1 ) - (O4 - O3 ) • Posttest-only control group design (after only with the control group): Experimental Group: R X O1 Control Group: R O2 Treatment effect: O1 - O2
Bivariate regression model
• Y = β0 + β1X + ε Y dependent (outcome) variable X independent (predictor) variable β0 intercept of the regression line β1 slope of the regression line ε error (the difference between actual and predicted values)