Comstrat 309 1nd EditionExam # 1 Study Guide Survey Design- Types of surveyso Descriptiveo Analytico Formativeo Summative- Survey compositiono Items Used to measure a phenomenon of interest Items can be continuous, nominal, and ordinal Continuous items can be averaged - Assumptions of a continuous survey scaleo A=b=co Semantic equidistance Difference between items is equal o Likert scale  Evaluate the statement Very dissatisfied- satisfied o Problems with neutral midpoint Not semantically equidistant Neutral is ambiguous to survey takers- Neutral can be neutral- It can mean “I don’t know”- It can mean “I don’t care” Bad idea for survey taking, not strong measurement validityo Open ended items Open-ended items often used in advertising research Give insight into what survey-takers are thinking Results can be quantified through coding - “What did you think overall?”- “What would you improve?”- Surveys good from bado Double barrel questions Asks more than one question, but assumes one answer Example- To what extent do you agree with these items:o Comstrat 309 is useful and funo I am angry and sad Greatly affects measurement validity - You don’t know if someone is just angry or just sado Suggestive survey questions Imply certain answer should be given Falsely presents presupposition in question as accepted fact o Leading questions When prior questions can influence the response to future question Leading questions prime responses to further questions - Politically charged type surveys o Push polls Appear as political polls Actually designed to persuade and promote certain ideology Contain suggestive and leading questions Survey Validity and ReliabilitySurvey informational bias due to invalid instrumentation- Semantic equidistance- Suggestive questions- Double-barrel questions- Leading questionsInformation Bias- Invalid instruments- Recall bias- Social disability response biasSelf report- Benefitso Quick responseso No need for behavioral measureso Can quickly compile data- Drawbackso Dunning-Kruger effecto Social desirability responseo Recall biasDunning-Kruger Effect- Perceived knowledge = actual knowledge?Remedy to Dunning-Kruger Effect- Ask items that tap into knowledge- Make sure these items are good operationalization’s of your target variableSocial Desirability Bias- Over/under-estimating response due to social pressures- Threatens measurement validity- Example: o Mental health questionnaireOvercoming Social Desirability Bias- Ensure participant confidentiality- Make survey anonymousRecall Bias- Often bad at recalling events- Sometimes false memories occur Remedy to Recall Bias- Make sure survey items aren’t suggestive or leading- Priming can lead to false recollectionso “Many families go on vacation to Disney World. As a child, how often did you go on vacation to Disney world?”System Item Reliability- Operationalization  Measurable Variable- Religiosity  o 1) Religious service attendanceo 2) Belief in supreme beingo 3) Donation to religious institution - Question: Does a survey respondent rate these three different items in a similar way?Survey Measuring Religiosity- Good reliability  same survey responded rates the following in similar wayso 1) Religious service attendanceo 2) Belief in supreme beingo 3) Donation to religious institution Survey analysis: the basics- Survey measuring religiosityo Good reliability- same survey respondent rates the following in similar ways - Analyzing reliabilityo Cronbach’s alpha= statistical test that determines internal consistency of items o If anything is above .7 then we would say we have a reliable set of numberso Anything below may question reliability - Reliability does not equal validity- Analyzing survey resultso Parametric statistics Assumptions Normal distributions and the probability that is assumed based of these distributions  U- sample mean 0- sample standard deviation- Amount of variation from the mean Large standard deviation- Wide distribution Small standard deviation- Peaked distribution- They can have the same means but different degrees in variationo Types of statistical tests T-tests- Used to test differences between two sample meanso Differences between two independent means- Exampleo Surveys attitudes about a producto Do men score differently than women - Assumptionso Data is normally distributedo Homogeneity of variance Bothe samples are of the same in variationo Independence Both sample means independent of one another o If variances scores were different (p <.05) then this assumption is violated Correlation- Relationships between one independent variable and one dependent variable - Linear relationship between x and y variables- Tests the relationship between those two variables- Positive correlationo As x goes up, so does y- Negative correlationo As X goes up, Y goes down- Data must be normal- The strength of a relationship is measured with Pearson’s r coefficient - P value determines significance- Case studyo Attitudes about diet water Intentions to purchase diet water- Often confused with causation Multiple regression- Relationship between multiple independent variable and one dependent variable- Same as correlation analysiso Allows for multiple independent variables instead of only oneo With multiple independent variables, you can control for confounding factors- B- standardized coefficiento Tells you how strong the relationship is between your X and Y variables- R-Squaredo The amount of variance explained by your independent variables in predicting DV Chi-square - Difference between the expected frequencies and the observed frequencies in ore or more categories o Only have categorical independent and dependent variables Cant be averaged Yes-no, male-female - IV- male/female- DV- purchase drink Y/NContent Analysis- Changes in Media Content?- Has media become more violent?- Is there greater representation of minority groups in media?- Definition:o A systematic procedure to examine the content of recorded informationContent Analysis: The Foundation- Systematic:o Content is analyzed using explicit and consistently applied rules 1) Sampling: Must follow explicit procedures for generalizability 2) Evaluation: All content evaluated in same manner- Objective:o Researchers biases