SU3- Qualitative Data, Descriptive Research

Qualitative data analysis

Characteristics

Steps

Quantitative (descriptive) data analysis- survey

Error types

Survey methods

Person-administered

Cons- slow data collection, recording errors, interviewer-respondent interaction error, expensive

Selection criteria

Focus- increase understanding (vs. quantifying magnitude or explaining) of phenomenon

Criticism- lacks rigour (words/text, images/visuals vs. numbers)

Qualitative vs. quantitative

Qualitative

Quantitative

Data- text/words, visual/images

Focus- understanding

Process

Member checking- key informants verifies report/analysis accuracy; solicit feedback

Ongoing, iterative- simultaneous data collection + analysis; affects future data collection (e.g. qns asked)

Inductive- data > themes; grounded theory

Data- numerical

Focus- quantifying (descriptive), explaining (causal)

Process- deductive i.e. themes (defined before collection) > data

(Preliminary) managing data collection- transcription, data entry (observations, notes, respondents' narratives)

Data display (sum/reduce data) e.g. tables (explains key themes), figures, flow diagrams (suggests variable r/s), trees, maps, matrix (quotes for various themes from key informants)

Conclusion drawing, verificatioon

Data reduction

Categorisation- group by similar content

Coding- (labels/numbers attached to different categories)- for tracking, organisation

Comparison (similarities/differences in categories identified)- helps refine theories

Integration (theory-building)

Data reduction

Tabulation (response count)- number of times selected categories mentioned

Code sheet- list different categories + respective coding

Recursive, circular r/s i.e. variable is both cause, effect

Selective coding- one core/central category (frame) storyline built around

Iteration (repetitively working thru data)- modify/revise earlier analysis by uncovering issues unaddressed by alr collected data

Negative case analysis (deliberately looking for instances contradicting developing theories)- establishes theory conditions

Most frequently coded/mentioned- implies importance, further investigation/analysis/follow-up

Don't report % (implies statistically projectable results)

Co-occuring themes (concepts frequently mentioned together)- r/s

Does not imply magnitude of finding- use quantitative research

Bias check

Establish credibility

Emic validity- key members within culture agree with report findings

Cross-research reliability- data coding similarity by different researchers

Triangulation- multiple perspectives analysis e.g. different data collection methods, analyses, sets, researchers, time periods

Peer review- by external specialist/methodology

First impression salience

Selectivity (overconfidence of some data)- confirmation bias

Co-occurrences- taken as correlation/causal r/s

Extrapolating instances rate to population- when it is non-generalisable

Not accounting unreliable info sources

Report writing

Intro

Data analysis, findings

Conclusion, recommendations (actionable implications)

Methodology description

Research qn, objectives

Lit. review

Data displays

Findings- interpretation, summary

Topics covered

Methods used, location, date, time

Researcher info- qty, background

Participant selection procedure

Participant info- qty, characteristics

Data analysis procedures e.g. coding, peer review, etc.

Limitations

Usage

Research problem nature- describe specific situational characteristics/evaluate strategies

Research qn- focuses on who, what, when, where, how

Task

Identify meaningful r/s

Determine true differences- magnitude

Verify r/s validity

Sampling error- finding differences (sample vs. population true value)

Non-sampling error (not-related to sampling; systematic)

Reduce

Increase sample size

Appropriate sampling method

Sources

Respondent

Measurement/questionnaire design

Problem definition

Project administration

Characteristics

Creates systematic variation/bias

Controllable

Cannot be directly measured

Creates other non-sampling error

Cannot be reached

Unwilling to participate

Inaccurate response

Types

Nonresponse error- final sample different from planner sample i.e. significant number of preselected prospective respondents refuse to participate

Response error- respondents impaired memory/inaccurate response

Pros- adaptable, rapport, feedback, response quality

Telephone-administered/interviews

Self-administered- respondent record own response (without interviewer' presence)

Types

In-home

Executive- typically in office with biz executive

Mall-intercept

Purchase-intercept0 intercept after product purchase/selection

Pros- cheaper, faster, more respondents, close supervision of interviewers, wide geographic area, callback for respondents who did not answer, random digit dialing (random sample)

Cons- cannot present non-audio stimuli, respondent hang up (esp. for lengthy interview), limited by national borders (rarely used in international research), high refusal rates, annoyed by interruption, poor telephone iinterview perception due to sugging (selling under research disguise)

Computer-assisted telephone interviews (CATI

Interviewer reads qns from screen

Enters ans directly into comp.

Pros- lower cost, can send inbound calls to interviewer at later time, eliminates separate data entry and edits, real-time results tabulation

Types

Mail

Mail panels- individuals agree to participate in advance

Drop-off- respondent completes and returns survey to researcher

Internet- online surveys

Pros- no interviewer cost (training, traveling), search cost, can reach hard-to-interview ppl

Cons- low response rate (nonresponse bias), misunderstood/skipped qns, slow (time-lag btw survey mailing, returning)

Pros- screen in advance (ensure representative sample), high response rate

Useful for longitudinal research (across periods) - observe changes over time

Pros- respondents' availability, screening

Cons- expensive

Pros- cheaper (than other survey methods), most frequently used, no coding needed, can reach hard-to-reach samples, techonology has improved functional capabilities over paper surveys, can randomise qn orders (remove effects of qn order)

Situation

Task

Respondent

Budget

Time frame- longer = higher quality data

Data quality requirements

Difficulty- complex survey (need researcher presence, conduct)

Stimuli (e.g. product, ads) to elicit response- telephone limits visual stimuli

Amt of info

Topic sensitivity

Diversity- degree of shared/common characteristics

Incidence rate- % of general population making up research focus

Participation degree

Ability- mail survey eliminates need for continuous time to complete questions, telephone survey can call back at more convenient timing

Willingness- uninterested, busy, privacy concerns

Knowledge lvl- more detailed info requirement, respondent to be more highly knowledgeable

Affected by scale measurement, questionnaire design, sample design, data analysis

Completeness (data depth, breadth)- full vs. vague picture

Generalisability- represents studied population

Precision (response accuracy)

Non-generalisable- low response rate, small sample size

Mail, internet- precise, not generalisable (hard to get representative sample)

Telephone- generalisable but not precise (short qns, interview duration)

Large, detailed info

Needs trained interviewer

Lower response rates

Higher respondent fatigue

Online (vs. phone, face-to-face) survey- more honest

Low- need more resources to get respondents

Goal

Reduce search time, cost

Maximising data amt