3: Descriptive research Flashcards
DR vs experimental research
describing a population OF INTEREST (w/o causality)
Vs
understanding causal effects
DR vs surveys
DR includes surveys (USUALLY PAID) n other methods
DR answers these 2 Qs:
- who are your customers / electors / employees?
- what are they doing / thinking in what %?
Data can be classified: - By collection purpose - By source
- primary (collected for this study) or secondary
- internal or external to your org
Designing a survey: 4 Qs = 1 what + 3 how’s
- what to ask? > Prioritize main Qs
- how much to ask? < Participants’ interests; Time after 5’ attention collapses
> good strat: ask filter Qs (causing branching!) < - how to phrase? > Make it simple (n fun)
- how to arrange Qs? > Order from most important (laddering?)
Designing a survey: Structure in 4=1+2+1 parts
- introduction (inform who is investigating what / why, how long, confidential)
- intro Qs: easy > make participants feel confident n committed + screening Qs: to ensure quotas for representation
- sensitive n related Qs: real focus of survey, often of sensitive n personal nature
- end study: thank n leave contact for further enquiry
Designing surveys: 4 aspects to consider reg. Qs (1+2+1)
- Q content: only relevant
- Q response format: closed or open-ended
> scaling
- Q response format: closed or open-ended
- Q wording > simple
- Q order > beware of interdependencies
Q content: 3-4 aspects to consider (1 + (1 + 1-2) )
- how many n which Qs? ~ complexity; consult research lit, eg Marketing Scales Handbook
- can respondents in yr audience answer yr Q?
> Include “I don’t know”
- can respondents in yr audience answer yr Q?
- Q needed, acceptable? Eg income
> use edu or other proxy
- Q needed, acceptable? Eg income
- will respondents answer truthfully? Social desirability > ask about others in situation
Q response format: 2 possibilities
- open > explanatory, but more difficult for stat analysis
- closed: neutral answer? (w odd-numbered answers)
> it can be better!
- closed: neutral answer? (w odd-numbered answers)
Q wording: 6=3+3 aspects to deal with
answerability:
- simple n audience-specific language
- avoid ambiguity (eg do you exercise regularly?)
- avoid burdensome Qs, instead ask memorable Qs w time referents
unbiasedness:
- avoid leading Qs (do not add info)
- split “double-barrelled” Qs in 2, as they should be
- instead of making assumptions (eg of relationships, causal or other, in the Q), split them
Q order: problem to solve
- interdependencies < consistency bias
> Separate w other / filler Qs!
Collecting data 4 aspects to treat: 2+2
design:
- who should participate?
> Set quotas (adjustment by mathematical weighting only works if you are close to target quotas) - how often? Cross-sectional vs longitudinal
user-friendliness:
- may response rate be low?
> Increase subjects’ interest - survey works well n in designed time?
> Pre-test it
Sampling: 2 issues to consider
- tradeoff: saves money vs asking whole population, but should be representative
- quotas help: eg gender according to demography …but you never know for sure what all aspects to include (use covariates n control for them)
Sample types:
- Cross-sectional
- multi-cross-sectional
- longitudinal
- panel
- 1 time 1 sample
- 1 time N samples
- N times same 1 sample
- Ongoing (good for trends but theres panel mortality…)
To increase response rate… (1 overarching theme) 4=2+2 pieces of advice
(market yr survey well!)
2 tactical:
- time invite / reminder well
- keep it short
2 motivational:
- mention why, credible affiliation, show who you are
- use rewards (lottery for cool prize works best)
