You are on page 1of 16

at

17/11/2009 15:07:00
Chapter 8

Measurement0 the systematic profess of quantifying information about constructs, concepts, or objects Measurement process consists of two tasks o Construct selection/development o Scale measurement Construct Development Construct- unobservable concept that is measured indirectly by a group of related variables o Constructs are made up of a combo of several related indicator variables that define the concept being measured Construct development begins with a definition of the purpose of the study and research problem Construct Development- process in which researchers identify characteristics that define the concept being studied by the researcher o Heart of construct is need to determine what is to be measured

Scale Measurement Scale Measurement- the process of assigning descriptors to represent the range of possible responses to a question about a particular object or construct o Descriptors are a combo of labels (Strongly agree, strongly disagree) Scale Points- designated degrees of intensity assigned to the response in a given questioning or observation method o Scale points an be numbers (10 important, 1 not), or definitely important, moderately important etc) o All scale measurements can be classified as one of four basic scale levels nominal, ordinal, interval, and ratio

Nominal Scales Nominal Scales- type of scale in which questions require respondents to provide only some type of descriptor as a response o The most basic and least powerful scale design o Can only calculate the mode Ex- do you like chocolate ice cream yes or no? Ordinal Scales Ordinal Scales-scale that allows a respondent to express relative magnitude between the answers to a question o More powerful than a nominal scale o Greater than/less than, determine relationships between responses o Cannot be used to determine difference between rankings o Mode and median Ie can determine if they prefer coke over pepsi but not how much more Interval Scales Interval scales- scale that demonstrates absolute differences between each scale point o In addition to mode and median, the mean and standard dev of respondents answers can be calculated o Can determine differences as well as absolute differences Ratio Scales Ratio Scales- scale that allows the researcher not only to identify absolute differences between each scale point but also to make absolute comparisons between raw responses Ratio scales generally ask respondents to provide specific numerical values as their response, regardless of wheter or not a set of scale points is used Builds on interval scales measures by adding comparison between levels along with mode, median, mean, and standard dev

Evaluating measurement scales All scales should be evaluated for reliability and validity o Scale reliability refers to extent to which a scale can reproduce the same or similar measurement results in repeated trials o Validity assesses wheter scale measures what it is supposed to measure

Developing Scale Measurements


Designing scales requires o Understanding defined research problem o Establishing detailed data requirements o Identifying and developing the constructs o Knowing that a complete measurement scale consists of three components the question, the attributes, and the scale point descriptors After the problem/data requirements understood, research must develop constructs Next appropriate scale format (nominal, ordinal, etc)

Criteria for Scale Development Understanding of the questions Discriminatory power of scale descriptors o Discriminatory power- the scales ability to differentiate between categorical scale responses Balanced versus Unbalanced Scales Forced or nonforced choice scales o A scale that doest not have a neutral descriptor to divide positive and negative answers is referred to as forced-choice scale

o Non forced/free choice includes a netural response Desired Measures of Central Tendency and Dispersion o Measures of central tendency- locate the center of a distribution of responses and are basic summary stats Mean, median, and mode measure central tendency o Measure of dispersion- sample statistics that allow a researcher to report the variablility of the data collected from scale measurements, they are the frequency, distribution, the range, and the estimated sample standard dev o Nominal scales can only be analyzed using frequency distribution and mode o Ordinal scales only medians and ranges, as well as modes and frequency disturb o Interval/ration scales, most appropriate statistic to use are means and standard deviations

Scales to Measure Attitudes and Behaviors


Likert Scale Likert Scale- ordinal scale format that asks respondents to indicate the extent to which they agree or disagree with a series of statements Initially had five descriptors: strongly agree, agree, neutral, etc Best for research designs that use self administered surveys, personal interviews, online surveys

Semantic Differential Scale Semantic Differential- unique bipolar ordinal scale format that captures persons attitudes or feelings about a given object o (good/bad, like/dislike, etc) Behavioral Intention Scale

Behavioral Intention Scale- special type of rating scale designed to capture the likelihood that people will behave in some way regarding a product or service o Good predictor of consumers choice sof frequently purchased and durable consumer products Easy to construct

Comparative and Noncomparative Rating Scales


Noncomparative rating scale- used when objective is to hav a respondent express his or her attitudes, behavior, or intentions about a specific object or its attributes Comparative Rating Scale- used when objective is to have a respondent express his or her attitudes feelings or behaviors about an object or its attributes on the basis of some other object or its attributes Graphic Rating Scales- use a scaling descriptor format that presents a respondent with a graphic continuum as set of possible response to a question Rank order scales- use a format that enables respondents to compare objects by indicating their order of preference or choice from first to last Constant Sum Scales- ask respondents to allocate a given number of points o Points are often allocated based on importance of product features to respondents

Other Scale Measurement Issues


Single-item and Multiple-Item Scales Single-Item Scale- involves collecting data about only one attribute of object or construct being investigated ie. Age Multiple-Item Scale- one that includes several statements relating to the object or construct being examined o Decision to use either is made when construct is developed o Multiple item scales more reliable and valid Clear Wording

Use clear wording and avoid ambiguity, avoid using leading words Screen Questions Purpose to identify qualified prospective respondents Skip Questions

Chapter 9
Values of Questionnaires in Marketing Research
Questionnaire- document consisting of a set of questions and scales to gather primary data

Questionnaire Design

Discussion of steps based on study conducted for American Bank, bank wanted to expand regionally, to improve decision making info needed on banking habits patterns satisfaction and commitment as well as demographic and lifestyle characteristics of current and potential customers 1) Confirm Research Objectives 2) Select Appropriate Data Collection Method 3) Develop Questions and Scaling 4) Determine Layout and evaluate Questionnaire 5) Obtain Initial Client Approval 6) Pretest, Revise, and Finalize Questionnaire 7) Implement The Survey

Step 1: Confirm Research Objectives 1) Collect data on selected demographic characteristics that can be used to create a profile of current and potential customers 2) collect data on selected lifestyle dimensions that can be used to better understand current customers and banking habits of potential 3)identify preferred banking services, as well as attitudes and feelings towards those services 4)identify demographic and lifestyle characteristics of market segments as well as satisfaction with and commitment to current

primary banking relationship Step 2: Select Appropriate Data Collection Method Step 3: Develop Questions and Scaling Question format (structure or unstructured), wording of and instructions for responding to questions and scales, type of data required (nominal, ordinal, interval), construct Unstructured Questions- open ended questions that allow respondents to reply in their own words o More difficult to code for analysis Structured Questions- closed ended questionst hat require the respondent to choose from Wording Questions and Scaling o Bad questions Unanswerable- because respondent does not have access to information or because none answers apply Leading or loaded- respondent is directed to a response that would not ordinarily be given if all responses were provided Double-Barreled- ask respondent to address more than one issue at a time do you drink pepsi with pbreakfast, olunch and dinner? Step 4: Determine Layout and Evaluate Questionnaire

Questionnaires begin with introductory section that gives respondent an overview of research Provides instruction on filling out survey Demographic questions at the end after comfort established Step 5: Obtain Initial Client Approval Duh Step 6: Pretest, Revise, and Finalize Questionnaire Pretest given to small representative group of respondents that are asked to fill out survey and provide feedback. Used to determine time Step 7: Implement the Survey

Chapter 12
Comparing Means: Independent versus Related Samples Researchers measure mean when means are from independent samples and when the samples are related o Independent sample ex.- results of interviews w/ male/fem coffee drinkers o Related samples ex.- when researcher compares average number of cups of coffee consumed by males with average number of soft drinks Using The t-Test to Compare Two Means Useful when the sample size is small (N<30) and when population standard dev is unknown Unlike univriate test, we assume samples are drawn from populations with normal distributions and that variances are equal T-test for differences between group means cab ne conceptualized as difference between means divided by variability of means

SPSS Application-Independent Samples t-Test

Chapter 11
Examining Relationships Between Variables
If a systematic relationship exists between two variables, then a relationship exists Linear relationship- strength and nature of the relationship between them remains same over range of both variables and can be described using straight line Curvilnear- strength and dirction of relationship changes over the range of both variables

Covariation and Variable Relationships


Covariation- the amount of change in one variable that is consistently related to the change in another variable of interest o If dvd purchases relate to age, we want to know extent to which younger persons purchase more DVDs Scatter Diagram- graphic plot of relative position of two variables using a horizontal and a vertical axis to represent the values of the respective variables

CHAPTER 13
Correlation Analysis
Pearson Correlation Coefficient- measures degree of linear association between two variables o Larger the correlation coefficient, stronger association between two variables o Null hypothesis for PCC states theres no association between two variables and correlation coefficient is zero

o Size of correlation coefficient can be used to quantitatively describe strength of association between two variablesWe use SPSS to consider statistic significance level(probability that null hypothesis will be rejected when in fact it is true) Pearson Correlation Coefficient We make several assumptions when calculating ^^^ o Two variables have been measured using interval or ratio scaled measures o Relationship being measured is linear o Variables you want to analyze have a normally distrubted population SPSS Application- Pearson Correlation

Substantive Significance of the Correlation Coefficient When coefficient is weak two possibilities considered o There is no consistent systematic relationship between two variables o Association exists but it is not linear and other types of relationships must be investigated further Coefficient of Determination- r2, number measuring the proportion of variation in one variable accounted for by another. The r2 measure can be thought of as a percentage and varies from 0.0 to 1.00 The larger the size of the coefficient of determination, the stronger the linear relationship between the two variables being examined Influence of Measurement Scales on Correlation Analysis Spearman Rank Order Corellation Coefficient- recommended statistic to use when two variables have been measured using ordinal scales o If either one of variables is repped by rank order data, spearman>Pearson to use in this situation

SPSS Application-Spearman Rank Order Correlation SPSS Application- Calculating Median Rankings

What is Regression Analysis?


Bivariate Regression Analysis- statistical technique that uses information about the relationship between an independent or predictor variable and a dependent variable to make predictions o Values of independent variable are selected and behavior of the dependent variable is observed using formula for straight line By comparing actual value (Yi) with predicted value (Y) we can tell how well our model predicts the actual value of our dependent variable Key assumptions o Regression analysis assumes linear relationship is good description of relationship between two variables o Regression analysis uses knowledge about level and type of association between two variables to make predictions o Regression model assumes 1) variables of interest are measured on interval or ation scales, 2) variables com from a normal population and 3) error terms associated with making predictions are normally and independently distrubted

Fundamentals of Regression Analysis Y=a + bX+ ei o Y= the dependent variable o A= intercept (pt where straight line intersects y-axis when x=0) o B= the slope (the change in Y for ever 1 unit change in X) o X= the independet variable used to predict Y o Ei=the errer for the prediction Least squares procedure- determines best fitting line by minimizing vertical distances of all data points from the line

Unexplained variance- is the amount of variation in the dependent variable that cannot be accounted for by the combination of independent variables o Unexplained variance is called error and is repped by vertical distance between estimated straight regression line and actual data points Distance of all points not on line are squared and added to determine sum of squared errors, which is measure of total error in regression

Developing and Estimating the Regression Coefficients Ordinary Least Squares- is a statistical procedure that results in equation parameters that produce predictions with the lowest sum of squared differences between actual and predicted variables Differences between actual and predicted values of y are respesented by ei, add them up, the total respresents overall measure of accuracy of regression equation

SPSS APPLICATION- Bivariate Regression CHECK Significance CHECK Multiple Regression Analysis Multiple Regression Analysis- technique which analyzes linear relationship between dependent variable and multiple independent variables by estimating coefficients for the equation for a straight line examine regression coefficient for each independent variable, which represents average amount of change expected in y given a unit change in value of independent variable standardized regression coefficient is called beta coefficient- shows the change in dependent variable for each unit change in the indepdent variable, standardization removes effects of using different scales of measurement o for example, years of age and annual income measured on different scales

positive beta means as size of an independent variable increases, size of the dependent variable increases. A Negative beta means that as the size of the independent variable increases, size of dependent gets smaller

Statistical Signficance After regression coefficients estimated, examine stat signifiance of each coefficient Done same as bivariate regression, each is divided by standard error to produce t stat, which compared against crit value to determine whether null hypothesis can be rejected When using multiple regression analysis, it is important to examine overall statistical significance of regression model Model F statistic, stat that compares amount of variation in the dependent measure explained or associated with independent variables to unexplained or error in variance. Larger F indicates the regression model has more explained variance than error

Substantive Significance The multiple r2 or multiple coefficient of determination describes the strength of the relationship between all independent variables in our equation and the dependent variable The larger the r2 measure, the more of the behavior of the dependent measure is associated with the independent measures we are using to predict it Higher values of r2 mean stronger relationships between independent variables and dependent measure Steps to follow in evaluating results of regression analysis o 1) Assess statistical significance of overall regression model using F stat and its associated probability o 2) Evaluate the obtained r2 to see how large it is o 3) examine the individual regression coefficients and their t statistics to see which are statistically significant o 4)look at the beta coefficients to assess relative influence SPSS Application- Multiple Regression

CHECK

17/11/2009 15:07:00

17/11/2009 15:07:00

You might also like