Chapters12-18 Flashcards
Target population
Aggregate of all the individuals whose characteristics are the focus of the marketing research project
- target market
- sampling units
Census
Act of measuring all the individuals in the pop.
- expensive
- time consuming
- technically demanding
- intrusive
- may miss individuals systematically–>bias
Sample
Group of individuals drawn from a pop.
-smaller than target market
Sampling frame
A procedure used to select the sample for the target pop
- a list of all the individuals in the pop
- a rule or set of directions to find all the individuals in the target market
Bias
Systematic error: error-producing factor that is stable and operates in a constant direction
- cannot be quantified or identified
Random error
Sampling error: a sample does not include all members, so it may underestimate or overestimate the true avg. measure of a characteristic in a pop.
Simple random sample
-a probability sample
A probability sample; each sampling unit has the same chance of being sampled, and selection is based on a lottery model
- expensive due to sampling frame
- statistically inefficient
Cluster sample
-probability sample
Population is divided into subsets, and process is repeated several times
- reduces sampling costs
- high sampling error due to less precision
Stratified sample
-probability sample
The population is divided into segments, and each stratum is sampled separately
- increases high statistical efficiency
- more costly, the more strata, the higher the cost
Convenience sample
-non-probability sample
Selection of sampled individuals is based on ease of access and convenience
-exploratory/used for pilot surveys
Judgment sample
-non-probability sample
Selection of sampled individuals is done by interviewer based on some criterion
-exploratory/used for pilot surveys
Quota sample
Non-probability sample
Selection of sampled individuals is based on the goal of including certain subgroups & the judgment of the interviewer
Population parameters
The mean & standard deviation used to determine the distribution of a measure
Sample mean
It is a variable & depends on the sample
- own distribution: range of possible values
- the pop. mean is the most likely value
- normal curve
- the standard error of the mean gives a measure of the variability of the sample mean
- standard error of the mean is inversely related to the sample
What to consider when determining sample size?
- remain within budget constraints
- we want a precise sample statistic
- reliable sample statistic for inference about the population
How do we determine sample size?
We rely on secondary information to learn about the population.
-variability of the target measure
Confidence intervals approach
Population mean=sample mean +/- margin of error [m = X +/- (z)(sX)]
- Desired level of precision=margin of error
- Needed info: sd. in the pop.
- Statistical input: Z score corresponding to a certain confidence level eg. 95%:Z=1.96, 90%=1.64, 99%=2.58
Size of the margin of error
- smaller margin of error- higher precision
- the level of confidence + the sd. error of the sample mean
Standard error of the sample mean
Depends on the standard deviation of the measure in the pop. & the sample size
Margin of error
(Z)(sX) = (Z)(S)/square root of n
-set the acceptable level of confidence, level of Z
What makes a questionnaire defective?
- incomplete
- blank responses for most measures or key measures
- unsatisfactory or blank responses must be reported in project report
What is the bottom line of the sample size?
- large variability- high sd= large sample
- precise estimate-smaller margin of error-larger sample
- Precise estimate can be obtained by having lower confidence- greater uncertainty
Statistical inference
The process of extrapolation: estimating beyond the original observation range the value of a variable through a hypothesis
Hypothesis Ha
A discovery that you hope to establish
H0- the status quo
The ignorance hypothesis
- implies that we get no new knowledge from this piece of research
- is never really accepted
If research hypothesis Ha is true and the sample data is consistent with Ha=
Correct decision.
Probability: Power of the test (1-b)
If research hypothesis Ha is true and the sample data is inconsistent with Ha=
Sample evidence leads to Type II error
Probability that this occurs: beta (b)
If the research hypothesis Ha is false, and the sample data is inconsistent with Ha
=sample evidence leads to correct decision
Probability: Confidence level (1-a)
Product-moment correlation coefficient (r)
Measures the strength of linear relationship bet. Two variables
- variables bet. -1 & 1
- > if r is close to 1, rel. is strong & pos.
- > if r is close to -1, rel. is strong & neg.
- > if r is close to 0, rel. is weak
Simple regression
Assumes one variable affects the other but not vise versa (X–>Y)
Y=a+bX=e
Multiple regression
Represents the rel. bet. A dependent variable Y and a set of independent variables (X1,X2,X3,…) with a linear function
Y=a+b1X1+b2X2+b3X3+…+e
