Field Work, Error Review, Data CollectionWorker CharacteristicsSuccess Depends On:- Work Orientationo Extended Self- Aptitude- Discipline- “Third Ear”- Command- Ability to Woo- Ethical Supervision Requirements:- Sampling Control- Control of Cheating- Quality of Control and Editing - Central Office Controlo Tabulate responses for important demographic characteristics andkey questionsEditing Responses:(1) Field Editing- Performed by the interviewer- Done as soon after the interview as possible- Look for blanks, inconsistencies- Able to remember details - Able to follow up if necessary (2) Office or Final Edit- Again look for blanks and inconsistences - Get hold of the interviewer while they are still in town Data Cleaning:Identify “Blunders”- 5’s and 6’s in a 0/1 variableMissing Data- Leave Blank- Mean substitution (“neutral” value) (Most Often selected) - Delete case or just those with missing pairs- DK/NA- Compare missing against other known answers Summary:- Identification Field- Backup Data Sets - Cleaning time is well spent- Luck- “ Coding” covered in the Questionnaire Design sectionTotal Error Random Sampling Error Non Sampling Error Response Error Non-Response ErrorResearcher Error Interviewer Error Respondent Error Non-Coverage Surrogate Information Error Respondent Selection Inability Error non-responseMeasurement Error Questioning Error Unwillingness Error -refusalPopulation Definition Error Recording Error > really respondent? -not athomeSampling Frame Error Cheating ErrorData Analysis ErrorError Components Sampling Error:- Difference between the statistic and sampling distribution- Is increasing the sample size the answer?Non-Sampling Error:- Non-observationalo Non-coverage Sampling frame, researcher/interviewer judgment o Non- response  Refusal - Quality of introduction, incentives available, scary informed consent, sponsor anonymity, confidentialityo Not at Home Schedule of Call- backs Busier Segments of the population- Observational o Data Collection errors Interviewer Bias based on background, psychological and behavioral factors, errors in asking, probing, recording and falsification- Office processing errorso Editing, Coding, Tabulation, Analysiso ?? Is a poorly written survey a “data collection” or “office” problemBasic AnalysisGoodness of Fit: - Compare “actual” with “expected”o Use chi- square Works even with nominal datao Make use of known values for the population / must have values from the populationo Ho: there are not significant differences between the observed and expected distributions- List of stores purchased for a mail questionnaireo Stores were classified by sales revenue by the list owner Number of mutually exclusive categories =ko Received 200 responses to the survey o Use Kolmogorov- Smirnov Test (Vodka)  Requires data that is at least ordinalo Again use/need known valueso Ho: There are not significant differences between the observed and expected distributions -