Lecture 3 Flashcards
Secondary data
Secondary data already exists within the company or is collected by third parties for purposes other than solving the problem at hand - A gov publications, books, newspapers, annual reports
Uses of secondary data
Uses of secondary data:
Providing info at a sufficient level of details nd quality for solving a problem
Preliminary stage for solving a problem with primary data
Source for new ideas that can be investigated further with primary data, support for the problem definition and formulation of hypothesis
Limitations of secondary data
Limitations of secondary data:
Data is incomplete because it was generally collected for a different purpose
Units of measure and level of detail of the data do not correspond to the requirements
No control over the process of data collection
Data is too old
Primary data
Primary data is data does not yet exist and must’ve collected by the researcher or third parties
Origin of data
Either observation route or question route
Observation route: with or without survey
With its with lab or field experiment
Without it’s human as observer with internet or apparatus
For question route: qualitative or quantitative
Qualatitive is in depth interview or focus group discussion
Quantitative data for structured survey with primarily closed questions, passive (reactive ) role of the respondents
Qualitative data
Qualitative data is unstructured / semi structured survey, active role of the respondents
Quantitative data
Quantitative for structured survey with primarily closed questionsC passive (reactive) role of the respondents
Types of primary data collection: questioning and qualitative data
Personal interviews: greater willingness to discuss sensitive topics by excluding others. No motivation for social role playing. No pressure by other people.
Dis: establishment of a relaxed and sympathetic atmosphere. Ability of the interviewer to ask intelligent follow up questions to identity the reasons for behaviours and opinions.
Primary research: focus groups ads and dis
Focus groups: discussions in small groups
Ads; creative interaction between participants
Generation of a large quantity of info in a short time
Costs ads compared with personal interviews
Dis: aggregation of opinions is diffuse
Limited options for efficient, computer based processing
Types of primary research: observations
Without surveys: documentation of the behaviour of respondents without direct influence of the researcher - explatory, casual, understanding behavioural processes, uncoveringof interrelationships
With surveys: documentation of the subconscious behavioural reactionair the respondent to the stimuli / casual testing of advertising campaigns, markets / lab regs of new products
Scale
Scale is a discrete or continuous space onto which objects are located according to measurwmnt rules
Measurement
Measurement is rules for assigning symbolisms to objects such that these either numerically represent the amount of characteristics, define whether the object falls into a certain category
Types of scales
Nominal: categorisation of others eg classification of gender, social stratum, mental status W measure in %
Methods like chi square test (contingency analysis)
Ordinal - ranking of object in order eg preference ranking of brands - metrics like median, rank, order,
Methods like ordered regression, conjoint analysis
Ratio - assignment of numerical values to objects, whereby a natural zero point exists eg weight, scale, price
Methods are all
Formative measurement
Formative indicators measure the reason for change of a construct - number of beers, time passed since last alcoholic drink
Reflective measurements
Reflective indicators measure the effects of the change of a construct that is not directly observable eg clarity of articulation, ability to walk straight, reaction speed
Single item scales
Single item measurement of a construct of with only one item eg willingness to buy scale
Ads:
Simple, direct measurement, more scales can be measured by one survey
Dis: restricted value if important facets of a construct are not taken into account
Risk of reduced reliability for measurements of a variable that cannot be directly observed
Multi item scales
Multi item scales of a complex construct with multiple facets eg satisfaction scale
Ads: measurement of a phenomenon that is directly observable, increased relibikirt of the measurements for reflective scales
Dis: increased costs, danger of incorrect convulsion if the scale is erroneously defined as format reflective
Higher statistical requirements on the development and use of the scales
Quality criteria of measurement:
Validity / reliability / generalisability
High reliability wit high validity all xes in the middle
High reliability with low validity all exes in outside circles but together
Low reliability with high validity in xes around circle but not grouped together
Low reliability with low validity is scattered around the circle
Generalisability. Scale can be used for measurement in different settings
Probability sampling
Probability sampling: random selection of persons, prob including each person in same sample
Ads: sample remains representative of the population, derivations from the true value in the population can be calculated (sampling error)
Applications: standard application of data collection
Non probability sampling
Non probability sampling: selection of people based on a non random process
Ads; fast and cost effective execution
Applications: exploratory projects, homogenous products
The three:
Snowball
Quota
Convenience
Probability sampling procedures
Probability sampling procedures:
Simple random sampling
Systematic random sampling
Cluster sampling
All 3 are random sample of members of groups or interviews (eg random selection of regional units) use for interviewing of individual groups yield cost ads
Simple random sampling
Simple random sampling is randomised selections of respondents by random generator, drawing from a bowl or other methods. Use it for relevant groups are sufficiently larger and equally easy.
Systematic random sampling
Systematic random sampling is to pick every unit in a process that can be considered random (eg every 10th person in store)
Use it as simple to implement and takes ad of randomness in the environment
Cluster sampling
Cluster sampling is randomness section for groups that have representative composition. Use it for groups that have representative composition.
Stratified sampling
Stratified sampling is probability sampling of various groups within the population. Use: cost of sampling decreased of various heterogenous groups exist.
Proportional sampling -
Proportions of groups in the sample correspond to those in the population. Use of the sizes in the sample should correspond exactly to the relative sizes within the population
Disproportional sampling
Intentional over / under weighting of groups in the sample. Overweighting is required for small groups in order to obtain info results within groups (drivers of sport cars)
Non probability sampling procedures:
Snowball sampling
Quota sampling
Convenience sampling
Snowball sampling
Snowball sampling is after completion of interview, respondent is asked to name other people within a small, specialised population. Use it if pop is small and difficult to reach
Quota sampling
Quota sampling is intentional selection of respondents so that quota ls for specific criteria (eg age gender income) that correspond to the population are met. Uses are for few criteria for representativeness are relevant
Convenience sampling
Convenience sampling is selection of respondents who can be reached quickly and at low costs (eg students, colleague)
Uses are for small sampling are sufficient for thepurpsoe of the study. It is used for pre testing questionnaires
