MAR4613 Exam 3 Study Guide o Running two versions of something to test out which is better Chapter 14 Sampling AB Testing Census Sample o A sample that includes everyone o A sample that doesn t include everyone o Information that may be flaw Two Types of Errors o Assume there is truth out there in the market o Truth Estimate Error o Non sampling Error Hard to quantify o Sampling Error Spend to bring error down Easier to quantify Beat back by brute force Accuracy vs Cost Target Population To Be Sampled o Who is our target population Or our market Same thing o Find missing pieces of information to make marketing decision better o Sampling Unit Our sampling unit depends on the nature of our objective Usually an individual person household the thing the entity Ex Radio avg trip time sampling unit the trip o Sampling Frame List of sampling unit needs to be a concrete list Never that easy usually have to scramble to figure it out Ex DMV everyone who owns a car list Ex Political Campaign a specific demographic list o Specify Who Is Excluded o Consider Convenience Doable Can solve problem with finite amount of Probability Samples o Simple Random All equal Guarantees the sample resembles the population n has the same probability has all the other n Pick out of the hat o Stratified Sampling Census Random Different groups to sample Small number of groups within each group we sample Ex Gender and age Ex 30 doctors 100 patients o Cluster Sampling Random Random Randomly include some groups So many groups o Systematic Sampling Every other person Ex every 25th person o Multistage Design Three or more levels of random Within a country state and city Nonprobability Sampling can t calculate o Judgmental Sampling Expert opinion They decide who s in the sample o Snowball Design Give a survey to you then you pass it to others Pass it along o Convenience Sampling Handy inexpensive Pick people because they are cheap available o Quota Sampling Minimum number of people in each group Go to the mall survey 100 males and 100 females The Problem of Nonresponse o If you want 1 000 people and 50 of the people don t participate then you have to plan to interview 2 000 people o The bigger n is then the smaller the error will be o The number of people who participate may be different than the ones who do not participate Chapter 15 Sample Size and Statistical Theory Ad hoc o Made up on the spot o Different groups considered segments because they react different to the four p s o Subgroup A group of a group Ex Men divided into young or old group o Rule of thumb 100 per group 20 50 per subgroup 100 people in each group when have series of groups o Not much logic behind it but it is practiced often Statistics o Sample formula o English symbols o x sample mean summarizes the center of the distribution central tendency measures ratio interval numeric makes statement about it makes sense with high level of measurements deviation summarizes the dispersion of the sample summarizes how far they are away from the center high s value high dispersion low s value low dispersion when s is bigger the shorter the bell less average when s is smaller the taller the bell more average makes statement about o s o p proportion tends to be bars and not a curved line used for nominal values makes statement about Nominal related Would you go to my restaurant Yes or no Parameters Ex 120 200 p 6 o Market unobservable o Somewhat theoretical hypothetical no formulas o Greek symbols o the market s average the market s deviation o o Sample Reliability the true probability of the market o Mini bell o Averages for the entire sample repeated sample size of n o Combine samples average of the averages o All of the x bars are not identical because sample is finite o Dispersion will be less o The bigger the sample size leads to a thinner or tighter bell o The smaller the sample size leads to a wider or spread out bell The Confidence Interval o Probability falls between two lines o As the sample size n gets bigger the confidence interval line gets smaller Extras o The essential problem in marketing research is the missing pieces of information o We need information about the market but unfortunately we can only retrieve information from the sample o As the sample size n gets bigger the more x bar looks like o Statistics describes the sample o The x axis is what we measure Chapter 16 Data Entry and Data Analysis Independent Variable o Often one of the four p s in marketing o Sometimes will be attributes of consumers o The input o The cause o Ex Ad or no ad Dependent variable o The effect o The output o The consequence o Attitude award knowledge liking marketing outcome Extra o Use chi square when have nominal variables o Use regression when have numeric variables interval ratio Chapter 17 Hypothesis Testing 4 Steps of Hypothesis Testing Create mutually exclusive hypothesis Ho and Ha Only one hypothesis is right both can t be right Ho is safer opposite of what we believe in Ha is what we root for more dangerous Generate evidence bearing on Ho and Ha Assess the probability of the evidence assuming Ho is true Reject Ho if the probability is small 05 or less 1 in 20 Logic to Follow When Doing Hypothesis Testing Believe the opposite of what we want to believe Hypothesis are not symmetrical one is more dangerous Allow to control probability of danger Confidence Interval Makes a statement about the market from the sample information More of a description Hypothesis Testing Is designed to make all or nothing decisions A yes or no decision Chi Square Test Two nominal variables Row Demographic or 4 p s variable Ex Male or female Column Market response outcome If probability is Small Reject Ho Confirm Ha Not small Fail to reject Ho Do not confirm Ha Errors Type 1 Error Type 2 Error Don t spend money when we should Spend money when we shouldn t The error we worry about As a researcher we control the probability and the type 2 error Can control either errors Can t make it zero but can make it minimal Chapter 18 More Hypothesis Testing Single Nominal Variable z o Z is calculated assuming Ho is correct o Make assumption about the parameter using o Ex Would you go to my restaurant To see if we will breakeven o Steps Create two mutually exclusive hypotheses Generate evidence bearing on Ho and Ha find z Asses the probability of the evidence assuming Ho is true Three possibilities We are not in the ballpark p o We are in the ballpark and probability is not small Fail to
View Full Document
Unlocking...