Chapter 10 Data Collection 0 truth error systematic OR random sampling error I II Types of Error a Sampling Error b Noncoverage Error c Nonresponse error d Response error e Office error Sampling Error a Sampling error affects projects that rely on samples drawn difference between results obtained for a sample and the results we would have obtained had we gathered information from the population If you don t use a probability sampling technique it s impossible to estimate the degree of sampling error b III Noncoverage Error a Noncoverage error of the defined target population in the sampling frame error due to the failure to include some elements i Sampling frame problem b Reduce by improve the quality of the sampling frame i Use updated information ii Ex use computerized dialing approaches to handle unlisted numbers for telephone surveys IV Nonresponse error a Nonresponse error some elements of the population that were selected and designated for the sample error from failing to obtain information from b Potential problem because it only occurs when those who respond are systematically different in some important way from those who do not respond c Two Main Sources of Nonresponse bias i Refusals refuse to participate in a study resulting because some designated respondents 1 Mail web based interviews least affective arises when respondents are not at home when ii Not at homes the interviewer calls 1 Callbacks attempt to reach people selected for sample by calling at different days and times V Response Error a Response error consciously or subconsciously to a survey item individual provides an inaccurate response b Error can stem from researcher respondent or both c Questions for Considering Ways Response Errors can Effect Responses i Does the respondent understand the question 1 If not a Skip the question b Answer the question based on interpretation 2 Pretesting the questionnaire with members of the relevant population assist with preventing response error ii Does the respondent know the answer to the question 1 Perform sufficient exploratory research and questionnaire pretesting to understand what population members are likely and not likely to know iii Is the respondent willing to provide the true answer to the question 1 Attitudes emotions influence responses 2 Perform sufficient exploratory research and questionnaire pretesting iv Is the wording of the question or the situation in which it is asked likely to bias the response 1 Leading questions must be avoided 2 Don t use loaded words 3 Watch for unstated alternative assumed consequences double barreled questions 4 Don t ask respondents for estimates 5 Prevent by a Keeping respondent interested and motivated Interviewer records respondent s response by b carefully writing down the person s answers to open ended questions or checking appropriate box with closed ended questions 6 Interview cheating ethically wrong a Commercial research firms follow up with respondents so ensure they have been contacted b Cheating making up the entire interview or a few questions error due to data editing coding or analysis errors Office Error a Office error i Reduced by VI VII 1 Exercising proper controls in data processing Total Error is Key a Manage total error not a particular type b Increasing sample size does increase sampling error but increases other types of errors VIII Calculating Response Rates a Response rate units divided by the number of eligible responding units in the sample number of completed interviews with responding i Shows assessment of potential influence of nonresponse error 1 Higher response rates fewer problems with ii Response rate serves as quality indicator of data collection nonresponse bias 1 RR CI E a RR Response Rate b CI Number of Completed Interviews with Responding Units c E Number of Eligible Responding Units in the Sample b Web Based and Mail Surveys i Determine usable questions completed ii Determine eligible response units iii RR CI CI R NAH 1 RR Response Rate 2 CI Number of Completed Interviews 3 R Number of Refusals 4 NAH Number of Not At Homes c Telephone Surveys no eligibility requirement i Categorize Attempted Contacts in 3 Groups 1 Complete Interviews 2 Refusals 3 Non at homes ii Completed Interview once the respondent provides answers for most of the survey items d Telephone Surveys with eligibility requirement i Include response units that are not members of the population being studied 1 Ex Store wants to know shoppers opinion on the new layout Telephone interview those in the test market city Provide a screening question Has any adult in this household visited Goldsmith s Gallery in the past 3 months ii Count 1 Completed interviews 2 Refusal interviews 3 Not at home 4 Number of ineligible response a Help adjust for likelihood of the people who refused to take the survey or who weren t at home wouldn t have qualified anyway iii Adjust response rate to account for ineligible by using the eligibility percentage 1 E CI CI IE a IE Number of Ineligible Interviewees 2 RR CI CI E R NAH a RR Response Rate b CI Number of Completed Interviews c E Percentage of Eligible Interviewees d R Number of Refusals e NAH Number of Not At Homes IX Improving Response Rates a Survey Length i Respondents don t like long surveys 1 Get tired bored lose focus speed through the survey ii Shortened surveys make a respondent more motivated and etc respondent b Guarantee of Confidentiality of Anonymity Important with sensitive tops i ii Ethically bound to hold confidentiality iii Don t give managers personal information if they express interest in a product service c Interviewer Characteristics and Training i Similar race age and gender between the respondent and interviewer provides better cooperation ii Develop an effective recruiting script 1 Contain purpose of the study so the respondents develop greater involvement and interest i Ex hand addressed envelopes hand written signatures on d Personalization cover letters etc e Response Incentives Incentives increase response rates i ii Monetary incentives have the greatest influence iii Lotteries for a prize are effective f Follow Up Surveys i Often respondents fail to participate 1 Reminders are sent out a Removes anonymity Chapter 11 Data Preparation for Analysis Tribulate frequency distribution one way Skewness o Positively skewed left high values median is unaffected o Negatively skewed right small values at the tail Women 25 25 50 50 50 Sum 100 o Confidence interval
View Full Document
Unlocking...