Marketing Research: SMQ II and Sampling - PDF

Marketing Research Lecture 5: SMQ II and Sampling Professor Landry  Group Presentations Group presentation dates: March 26 or April 2 Email me by Friday (Feb 7) if your group would prefer to present on March 26, otherwise I’ll assume your group would prefer April 2...

Marketing Research Lecture 5: SMQ II and Sampling Professor Landry  Group Presentations Group presentation dates: March 26 or April 2 Email me by Friday (Feb 7) if your group would prefer to present on March 26, otherwise I’ll assume your group would prefer April 2 Will give priority to smaller groups; otherwise presentation dates will be randomly assigned Group report guidelines posted to quercus  =RAND() If you type this into a cell in MS excel and hit enter, it will generate a random number between 0 and 1 0.2761048804298… Morwitz et al. (1993): Dependent variable: likelihood of purchasing a new car within 6 months of survey (researchers got access to this data too!) Independent variable: whether or not person took the survey Treatment: asked question on left Control group: not asked question  Mere Measurement Effects New car purchases in next 6 months nearly 40% higher in treatment group (Morwitz et al., 1993, Journal of Consumer Research) Other mere measurement effects: Being asked about buying a PC → 20% higher rate of purchases (same study) Being asked about voting increases voting (Granberg & Holmberg, 1992) …in repeated online grocery purchases (Chandon et al., 2004), health insurance demand (Zwane et al., 2011), flossing (Levav & Fitzsimons, 2006) … “socially desirable behaviors” such as volunteering, recycling, blood donations as well as “socially undesirable behaviors” such as illegal drug use, stealing (Spangenberg et al., 2003; Williams et al., 2004)  Mere Measurement Effects: Repetition Morwitz et al. also tried repeatedly asking the same question to some of those who were asked about buying a product Find a “polarizing” effect: 1. Among those who said they planned to buy a new car: actual purchases in the next 6 months were over 10% higher if repeatedly surveyed vs. surveyed just once 2. Among those who said they never planned to buy a new car: actual purchases in next 6 months were almost 70% lower if repeatedly surveyed vs. surveyed just once  Mere Measurement Effects: Implications Even without “question bias,” asking questions can create bias by influencing the property you’re trying to measure Undermines representativeness or construct validity? Do responses account for effect of measurement? Counteract via decoy alternatives? Product category? Step 8: Determine Sample Plan and Size Population: who you want to learn about Doesn’t have to be individual consumers (e.g. could be stores) Sample plan: how each sample element, or unit, is to be drawn from the total population Goal: representativeness! Sample size: how many population units will be included in the sample Goal: accuracy! *alternative to sampling: conduct a census of entire population Suppose UTM is considering buying a vending machine that sells magazines for $2 each, and would like to collect information (via a survey) from its undergraduate students regarding… Whether students would be likely to use the vending machine if it were installed on campus Where the vending machine should be located What magazines should be made available  Suppose UTM is considering buying a vending machine that sells magazines for $2 each, and would like to collect information (via a survey) from its undergraduate students regarding… Whether students would be likely to use the vending machine if it were installed on campus Where the vending machine should be located What magazines should be made available What is the relevant “population” in this example? Should UTM draw a sample or perform a census? Suppose UTM is considering buying a vending machine that sells magazines for $2 each, and would like to collect information (via a survey) from its undergraduate students regarding… Whether students would be likely to use the vending machine if it were installed on campus Where the vending machine should be located What magazines should be made available If I (Prof. Landry) was the person responsible for administering the survey, would it make sense for me to just give the survey to everyone in this class?  Suppose UTM is considering buying a vending machine that sells magazines for $2 each, and would like to collect information (via a survey) from its undergraduate students regarding… Whether students would be likely to use the vending machine if it were installed on campus Where the vending machine should be located What magazines should be made available …in what ways would the students in this class not be representative of the population? …and how could this bias the results?  Basic Concepts in Sampling Sample frame: master list of population Sample frame error: degree to which sample frame fails to match the true population Incomplete vs. inaccurate Sampling error: any error that arises because a sample is used Due to sampling method or sample size  Types of Sampling Methods 1. Probability samples: members of the population have a “known” probability of being selected into the sample 2. Nonprobability samples: members of the population have an unknown probability of being selected into the sample  Probability Samples: Simple Random Simple random sampling: probability of being selected into the sample is known and equal for all members of the population Random device method uses an apparatus to ensure every member of the population has the same chance of being selected Old-school devices: coin, dice, deck of cards… Would flipping a coin be a practical way to draw a random sample from the MGT 453 sample frame?  Probability Samples: Simple Random Simple random sampling: probability of being selected into the sample is known and equal for all members of the population Random device method uses an apparatus to ensure every member of the population has the same chance of being selected Possible devices: coin, dice, deck of cards… Random numbers method: computer generates random numbers more practical when sample frame is electronic, for larger populations  Pilot Surveys You will be asked to summarize/tabulate the results of the pilot survey in the appendix of the final group project report Also include copy or screenshot of pilot survey You should review the pilot survey results before you design the final survey that you will give out – reviewing the pilot survey results will help you design better survey questions Final survey should be significantly different than the pilot Cluster sampling: researcher divides population into “clusters” that could represent population and confines study to some (randomly-selected) clusters One-stage: perform census of each selected cluster Two-stage: draw sample from each selected cluster Area sampling: use geographic areas as clusters Heterogeneous clusters → cluster specification error  Stratified sampling: population separated into distinct strata (subgroups), then a probability sample drawn from each strata Proportionate stratified sample: each strata’s proportion in the sample is roughly equal to its proportion in the population Otherwise it is a disproportionate stratified sample 3 5 1 4 2 5 4 1 5 3 3 Sample 1 2 4 5 2 3 5 4 1 3 2 4 5 5 Population 1 2  Stratified Sampling: Why & When to Use? Use stratified sampling when there is reason to believe there are substantial differences across easily-identifiable strata Graphical indicators: a non-bell-shaped distribution for key properties, e.g. skewed population, multi-peaked population Why stratified sampling? Allows for explicit analysis of each stratum Can improve accuracy & statistical efficiency  Stratified Sampling: Why & When to Use? Use proportionate stratified sampling if extent of heterogeneity within strata is believed to be roughly the same across strata Improves accuracy by ensuring each strata is adequately represented when using smaller sample size Otherwise: disproportionate stratified sampling Sample should draw disproportionately more from highly-heterogeneous strata than from strata that are relatively homogeneous Improves statistical efficiency by allocating more sample size to strata which require more data points to reliably estimate mean response Types of Nonprobability Samples Convenience: drawn at the researcher’s convenience, e.g. from a high-traffic area on a sidewalk or at a mall Purposive: researcher uses judgement or “educated guess” to decide who should represent the population Referral: ask initial respondents for names and contact information of other potential respondents Quota: researcher specifies percentages of the total sample for various types of individuals to be surveyed