BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

An Introduction to the Analysis of Surveys This handout is a brief introduction to the analysis of surveys. The handout walks you through the conceptualization of a model, descriptive statistics, reliability, discriminant validity, Regression, Analysis of Covariance (ANCOVA), Structural Equation Modeling (SEM), and Partial Least Squares (PLS). Most topics are presented briefly and somewhat superficially. For more details, please refer to individual courses that are part of your doctoral program in business, as well as related textbooks and published articles. The specific study presented in this illustration is from Evangelopoulos et al. (2003). The research study was motivated by a shortage in quantitative studies making a case for learning benefits of service-learning. Service-learning is an educational practice in which students do a project that serves an external entity, typically a non-profit organization and occasionally a governmental or industry unit. As the students get a chance to practice what they are learning in their course, they also provide some service to their community. The study examined university students taking a business statistics course that included a mandatory project, which could be research-oriented or service-oriented. Effects of project type on student perceptions were examined using a survey.

Conceptualizing a Theoretical Model Theory generation requires (i) a good literature review, (ii) intelligent observations, (iii) solid logical reasoning, and (iv) creative synthesis. Starting with the literature review, as the authors were looking for a model that would theorize relationships among course perceptions, they considered the intention to use the technique taught in the course as a core indication of the perceived value of the course. Marketing researchers are interested in a consumer’s intention to buy a product. IT researchers are interested in a user’s intention to use a technology. Thinking along these lines, a student’s intention to use the technique taught in a course in the future was selected as an important target construct. The literature review thus examined the theory of planned behavior (Ajzen & Fishbein 1980) and the technology acceptance model (Davis 1989). The Extended Technology Acceptance Model (TAM2), adapted from Venkatesh & Davis (2000) and shown below, offers a good starting point in understanding how a service-learning project might affect student perceptions.

1

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

In TAM2, result demonstrability was identified as an antecedent of perceived usefulness of a technology. Projecting this idea onto a service-learning context, we can associate involvement in a service-learning project, where the students demonstrate the results of the technique they learn in the class to an outside audience, with perceived usefulness of the technique taught in the course. Thus, the study introduces the Course Acceptance Model (CAM) as applied to the teaching of statistical techniques in a statistics course. The effect of involvement in a servicelearning project is theorized as an antecedent of perceived usefulness of the statistical techniques taught in the course.

Research Design A longitudinal study involving undergraduate students at a medium size public university in the Western U.S. was conducted. The students were enrolled in six sections of a business statistics course taught by three instructors. The course included a term project requirement. The project could be service-learning or research-oriented. At the risk of weakening the effect of servicelearning, course instructors avoided emphasizing the benefits of one type of project over the other. Workload and format of all graded project components were also kept equal for both types of projects. In order to overcome threats to internal validity generally associated with quasi-experimental designs, a control group design with a pre-test and a post-test was used (Cook & Campbell, 1979), with the students working on research projects acting as the control group. While one may argue that students working on research projects cannot be technically considered a no-treatment control group because they are exposed to an alternate treatment, such design allows to isolate effects of service-learning and to separate them from effects of other types of projects requiring application of course material. Perceived usefulness and ease of the course material, as well as attitudes towards the course and the intentions for future use were measured using a Course Acceptance Survey developed on the basis of the survey used to measure corresponding constructs in the original TAM (Davis, 1989). The questions of the questionnaire were tailored to the specifics of the course and included references to particular techniques taught in the course. The full text of survey questions is presented in the next section, together with the descriptive statistics. In order to capture the change in the students’ perceptions and attitudes over time, the survey was administered twice, in the middle of the semester (pre-test) and at the end of the semester (post-test). A passage of a significant period of time between a pre-test and a post-test allowed to minimize threats to internal validity due to testing effect (Cook & Campbell, 1979).

2

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

FOR MORE ON RESEARCH DESIGN: BUSI 6450 (research methods) BUSI 6480 (research design & ANOVA)

Descriptive Statistics Survey responses were captured using a 7-point Likert scale, where 1 = Strongly Disagree, and 7 = Strongly Agree. The sample sizes were N = 180 (Pre-Test survey) and N = 159 (Post-Test survey). The mean responses and the standard deviations (in parentheses) for each item are shown below.

Construct Measure Ease of Statistics (EA) EA1 Learning the Multiple Regression technique was easy for me EA2 It was easy for me to become skillful at using Statgraphics and performing statistical analyses EA3 Learning a number of Statistical techniques was easy for me Usefulness of Statistics(US) US1 Using statistical software and Statistical techniques would increase my performance in a business organization US2 Using Statistical techniques would enhance my decision-making skills as a manager US3 I would find Multiple Regression a useful tool that would enhance my problem-solving capabilities as a business consultant US4

Using Statistical techniques would enhance my decision-making skills as a manager

Attitude towards Statistics (A) A1 Do you enjoy performing statistical analysis? A2 Do you like statistics? Intention for Future Use (FU) FU1 How likely is it that you will be using statistics in the future? FU2 How frequently do you intend to use statistical software and statistical techniques in the future? FU3 Are you going to use data analysis and statistical modeling to support the future decisions of yourself or your organization?

Pre-test Post-test Mean (SD) Mean (SD) 3.97

(1.45)

4.73

(1.35)

4.74 4.27

(1.50) (1.46)

4.94 4.47

(1.46) (1.38)

4.97

(1.64)

5.12

(1.54)

5.13

(1.54)

5.13

(1.38)

4.62

(1.58)

5.04

(1.48)

*

*

5.08

(1.34)

3.58 3.65

(1.77) (1.74)

4.18 4.00

(1.59) (1.66)

4.18

(1.76)

4.42

(1.64)

3.73

(1.59)

3.96

(1.59)

4.08

(1.63)

4.16

(1.71)

You can replicate the descriptive statistics using data file CAMdata01.xls as described below: Using SPSS: 1.1. Start SPSS. Select File > Open > Data > files of type=Excel. Open CAMdata01.xls. Select Analyze > Descriptive Statistics > Descriptives. Select the 11 “Pre” variables. Click OK.

3

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Using Minitab: 1.2. Start Minitab 16. Select File > Open Worksheet > files of type=Excel. Open CAMdata01.xls. Select Stat > Basic Statistics > Display Descriptive Statistics > Descriptives. Select the 11 “Pre” variables. Click OK.

Reliability Internal consistency across different survey items (variables) attempting to measure the same construct is typically measured using Cronbach’s alpha, when the responses are on a Likert scale. For categorization studies using multiple raters use Cohen’s kappa (for two raters) or Fleiss’ kappa (for more than two raters). The measurement scales of the Course Acceptance Survey exhibited high reliabilities, with Chronbach’s alpha values ranging between 0.89 and 0.94. Pretest constructs (N = 180) as well as post-test constructs (N = 159) are listed below. Pre-survey and post survey items were identical, with the exception of a fourth usefulness item (US4) added to the post-survey, which was a repetition of US2.

Survey Items Pre Post 3 3 3 4 2 2 3 3

Construct Ease Usefulness Attitude Future Use

Reliability Cronbach's Alpha Pre Post 0.9033 0.8881 0.9111 0.9440 0.8989 0.9113 0.9291 0.9291

You can replicate the reliability statistics using data file CAMdata01.xls as described below: Using SPSS: 2.1. Select Analyze > Scale > Reliability Analysis. Select a group of variables corresponding to the same construct, e.g. PreEA1-PreEA3. Click Statistics. Check Descriptives for = Scale if item deleted. Select Continue > OK. Reliability Statistics Cronbach's Alpha

N of Items .903

PreEA1 PreEA2 PreEA3

Scale Mean if Item Deleted 9.03 8.28 8.72

3

Item-Total Statistics Scale Variance Corrected Item-Total if Item Deleted Correlation 7.745 .789 7.333 .801 7.421 .834

Cronbach's Alpha if Item Deleted .877 .868 .839

The results show that the reliability using all three items is equal to 0.903. Deleting any of the three items would result in reliability reduction. Therefore, we should keep all three.

4

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Using Minitab: 2.2. Select Stat > Multivariate > Item analysis. Select variables PreEA1-PreEA3. Click OK. The results are shown below. Cronbach's Alpha = 0.9033 Omitted Item Statistics Adj. Adj. Omitted Total Total Item-Adj. Variable Mean StDev Total Corr PreEA1 9.028 2.783 0.7886 PreEA2 8.281 2.708 0.8011 PreEA3 8.725 2.724 0.8344

Squared Multiple Corr 0.6269 0.6490 0.6964

Cronbach's Alpha 0.8775 0.8678 0.8389

FOR MORE ON RELIABILITY: BUSI 6450 (research methods)

Construct Validity and Confirmatory Factor analysis (CFA) Convergent validity examines whether items (variables) that are expected to be related to each other (because they attempt to capture the same construct) are indeed related. Discriminant validity examines whether items that are expected to be unrelated to each other are, in fact, unrelated. One way to examine convergent and discriminant validity is through confirmatory factor analysis. Using SPSS: 3.1. Select Analyze > Dimension Reduction > Factor. Select all groups of variables corresponding to the constructs examined in the study: PreA1-PreA2, PreEA1-PreEA3, PreUS1PreUS3, PreFU1-PreFU3. Click Extraction. Check Fixed number of factors = 4. Select Continue. Click Rotations. Check Varimax. Select Continue > OK. The results are: Rotated Component Matrixa Component 1

2

3

4

PreA1

.296

.299

.269

.810

PreA2

.246

.159

.289

.870

PreEA1

.835

.246

.201

.152

PreEA2

.865

.170

.183

.182

PreEA3

.837

.240

.173

.270

PreUS1

.255

.823

.315

.115

PreUS2

.231

.867

.248

.210

PreUS3

.241

.749

.349

.253

PreFU1

.224

.271

.843

.240

PreFU2

.194

.291

.859

.250

PreFU3

.226

.432

.737

.234

Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. a. Rotation converged in 6 iterations.

5

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Using Minitab: 3.2. Select Stat > Multivariate > Factor analysis. Select all groups of variables corresponding to the constructs examined in the study: PreA1-PreA2, PreEA1-PreEA3, PreUS1-PreUS3, PreFU1-PreFU3. Enter Number of factors to extract = 4. Select Type of Rotation = varimax. Select Continue. Click Rotations. Check Varimax. Select Continue > OK. The results are shown below. Rotated Factor Loadings and Communalities Varimax Rotation Variable PreA1 PreA2 PreEA1 PreEA2 PreEA3 PreUS1 PreUS2 PreUS3 PreFU1 PreFU2 PreFU3

Factor1 0.296 0.246 0.835 0.865 0.837 0.255 0.231 0.241 0.225 0.194 0.226

Factor2 -0.299 -0.159 -0.246 -0.170 -0.240 -0.823 -0.867 -0.749 -0.271 -0.291 -0.432

Factor3 0.269 0.289 0.201 0.183 0.173 0.315 0.248 0.349 0.843 0.859 0.737

Factor4 0.810 0.870 0.152 0.182 0.270 0.115 0.210 0.253 0.240 0.250 0.234

Communality 0.905 0.925 0.821 0.843 0.860 0.854 0.910 0.806 0.893 0.923 0.835

Variance % Var

2.6080 0.237

2.5968 0.236

2.5346 0.230

1.8368 0.167

9.5762 0.871

Output from SPSS and Minitab verify that we have good construct validity: Looking at the rotated loadings for the 4 requested factors we observe that all items related to the same construct have high loadings on one of the 4 factors, and all items unrelated to a given construct have low loadings on that same factor. The conclusion is that there is no need to drop any survey item and there is no need to re-conceptualize any of the constructs. FOR MORE ON CONSTRUCT VALIDITY AND FACTOR ANALYSIS: BUSI 6450 (research methods) BUSI 6240 (multivariate statistics)

Regression analysis After ascertaining that all four constructs have acceptable reliability and validity statistics for the four constructs, estimates for the four latent constructs were obtained by averaging answers to the items related to each construct. Variables Attitude1, Ease1, Useful1, Future1 (pre-test) and Attitude2, Ease2, Useful2, Future2 (post-test) were thus created. These calculated derived variables are already included in data file CAMdata01.xls. At this point the study focuses on a certain subset of the proposed model, the effect of project type (variable PostTYPE) on Useful2. Section and Useful2 were used as controls. In order to fit a regression model, a set of 5 binary variables are used to code the 6 different sections (variables SEC1-SEC5, already included in CAMdata01.xls). Since there was no particular reason for expecting an interaction effect, interaction terms were not included in the model. 6

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Using SPSS: 4.1. Select Analyze > Regression > Linear. Select Useful2 as the Dependent Variable. Select Useful1, PostTYPE, and SEC1-SEC5 as the Independent Variables. Click OK. The results are shown below. Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate 1 .676a .457 .420 1.025 a. Predictors: (Constant), SEC5, PostTYPE, Useful1, SEC3, SEC4, SEC2, SEC1 Model 1

Regression Residual

Sum of Squares 90.157 107.119

ANOVAb df 7 102

197.276

109

Total

Mean Square 12.880 1.050

F 12.264

Sig. .000a

a. Predictors: (Constant), SEC5, PostTYPE, Useful1, SEC3, SEC4, SEC2, SEC1 b. Dependent Variable: Useful2

Model 1

(Constant)

Coefficientsa Unstandardized Coefficients Stand. Coef. B Std. Error Beta 2.745 .398

Useful1 .535 PostTYPE .580 SEC1 -.989 SEC2 -.512 SEC3 -1.132 SEC4 -.339 SEC5 -.289 a. Dependent Variable: Useful2

.067 .266 .383 .300 .344 .320 .357

.597 .193 -.267 -.155 -.282 -.091 -.067

t 6.894

Sig. .000

8.033 2.180 -2.585 -1.706 -3.287 -1.057 -.810

.000 .032 .011 .091 .001 .293 .420

Since some coefficients are insignificant, the model is refitted after dropping them one at a time. The final model shows that Sections 2, 4, 5, and 6, form a group where perceived usefulness of the statistical techniques taught in the course is highest. Section 1 has a slightly lower average perceived usefulness and section 3 has an even lower average. The effect of doing a service project (PostTYPE = 1) is positive on perceived Usefulness, but only marginally significant. The final model has an R-squared of 44.1% and the following coefficient estimates:

Model 1

(Constant)

Coefficientsa Unstandardized Coefficients B Std. Error 2.413 .347

Useful1 PostTYPE SEC1 SEC3 a. Dependent Variable: Useful2

.551 .485 -.654 -.854

Stand. Coef. Beta

.066 .260 .326 .299

.615 .161 -.176 -.212

7

t 6.952

Sig. .000

8.392 1.863 -2.006 -2.852

.000 .065 .047 .005

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Using Minitab: 4.2. Select Stat > Regression > Regression. Select Useful2 as the Response Variable. Going directly to the final model (see SPSS instructions above), select Useful1, PostTYPE, SEC1 and SEC3 as the Predictors. Click Storage. Select residuals and fits, then click OK > OK. The regression analysis results are shown below. Regression Analysis: Useful2 versus Useful1, SEC1, SEC3, PostTYPE The regression equation is Useful2 = 2.41 + 0.551 Useful1 - 0.654 SEC1 - 0.854 SEC3 + 0.485 PostTYPE 110 cases used, 5 cases contain missing values Predictor Constant Useful1 SEC1 SEC3 PostTYPE

Coef 2.4134 0.55104 -0.6539 -0.8536 0.4851

S = 1.02498

SE Coef 0.3472 0.06566 0.3259 0.2993 0.2603

R-Sq = 44.1%

T 6.95 8.39 -2.01 -2.85 1.86

P 0.000 0.000 0.047 0.005 0.065

R-Sq(adj) = 42.0%

Select Stat > Basic Statistics > Normality Test > Variable = RES1 > OK. The normality assumption is supported (p-value = 0.167). Select Graph > Scatterplot > Simple > OK > Y variable = RESI1, X variable = FITS1 > OK. The normal probability plot and the residual plot, shown below, exhibit support for the model assumptions. Probability Plot of RESI1

Scatterplot of RESI1 vs FITS1

Normal

3

99.9

Mean StDev N AD P-Value

99

2 1

80 70 60 50 40 30 20

RESI1

Percent

95 90

-2.02666E-15 1.006 110 0.535 0.167

0 -1 -2

10 5

-3

1 0.1

-4 -4

-3

-2

-1

0 RESI1

1

2

3

2

3

4

5

6

7

FITS1

FOR MORE ON REGRESSION: BUSI 6220 (regression analysis)

Analysis of Covariance (ANCOVA) Analysis of Variance (ANOVA) and Analysis of Covariance (ANCOVA) are least-squares modeling approaches that are very similar to least squares regression modeling. We start by illustrating a design validation test. Recall that the research design is quasi-experimental. A true experimental design would randomly assign students to project type across all participating

8

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

sections (rather than allowing them to select a project by themselves). The fact that students were allowed to choose the project type has the following validity threat: What if those students who were already more convinced of the usefulness of the statistical techniques showed a higher preference for service projects? To investigate this question, we fit an ANCOVA model that explains Useful1 (i.e., perceived usefulness at pre-test) using PostTYPE (i.e., project type selected after pre-test) as an explanatory factor, after also accounting for Section. In order to fully account for Section, we include in the model Section, as well as the interaction between section and project type. Using SPSS: 5.1. Select Analyze > General Linear Model > Univariate. Select Useful1 as the Dependent Variable. Select PostTYPE as a Fixed Factor, and Section as a Random Factor. Section is a “random” factor because the 6 sections participating in the study are simply examples of possible sections. If we replicate the study we are likely to use different sections. However, the two project types are not two examples of project types. We are specifically interested in these two particular project types, service project and research project. If we replicate the study we will definitely include the same two project types. Therefore, project type is a “fixed” factor. Click OK. The results validate our design, by showing no significant effect of section, project type, or their interaction, on perceived usefulness. The results are shown below. Tests of Between-Subjects Effects Dependent Variable:Useful1

Hypothesis Error

Type III Sum of Squares 1164.544 16.329

1 23.572

Mean Square 1164.544 .693a

F 1681.037

Sig. .000

PostTYPE

Hypothesis Error

.160 16.447

1 7.240

.160 2.272b

.070

.798

Section

Hypothesis Error

1.844 11.328

5 5

.369 2.266c

.163

.966

PostTYPE * Section

Hypothesis Error

11.328 225.661

5 98

2.266 2.303d

.984

.432

Source Intercept

df

a. .832 MS(Section) + .168 MS(Error) b. .832 MS(PostTYPE * Section) + .168 MS(Error) c. MS(PostTYPE * Section) d. MS(Error)

These results imply that the students picked projects without regard for the effects of project type (service project versus research project) at the time of pre-test survey. Let us now examine how perceived usefulness had formed by the time of post-test survey. Continuing the analysis, fit an ANCOVA model for the effects of project type and section on perceived usefulness at post-test, using usefulness at pre-test (Useful1) as a covariate. Select Analyze > General Linear Model > Univariate. Select Useful2 as the Dependent Variable. Select PostTYPE as a Fixed Factor, Section as a Random Factor, and Useful1 as a Covariate. Select Model > Custom. In the Model panel (right panel) list PostTYPE, Section, and Useful1 as the model terms. Select Type = Main Effects. Click Continue > OK.

9

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Tests of Between-Subjects Effects Dependent Variable:Useful2

Hypothesis Error

Type III Sum of Squares 54.090 83.296

1 69.275

Mean Square 54.090 1.202a

F 44.985

Sig. .000

PostTYPE

Hypothesis Error

4.992 107.119

1 102

4.992 1.050b

4.754

.032

Section

Hypothesis Error

14.010 107.119

5 102

2.802 1.050b

2.668

.026

Useful1

Hypothesis Error

67.771 107.119

1 102

67.771 1.050b

64.532

.000

Source Intercept

df

a. .087 MS(Section) + .913 MS(Error) b. MS(Error)

Using Minitab: 5.2. Select Stat > ANOVA > General Linear Model. Select Useful2 as the Response. In the Model panel, list PostTYPE, Section, and Useful1 as the Model terms. List Section as a Random Factor. Select Covariates. List Useful1 as a Covariate. Click OK > OK. Click Continue > OK. The analysis results are shown below. General Linear Model: Useful2 versus PostTYPE, Section Factor PostTYPE Section

Type fixed random

Levels 2 6

Values 0, 1 1, 2, 3, 4, 5, 6

Analysis of Variance for Useful2, using Adjusted SS for Tests Source PostTYPE Section Useful1 Error Total

DF 1 5 1 102 109

S = 1.02479

Seq SS 0.196 22.190 67.771 107.119 197.276

Adj SS 4.992 14.010 67.771 107.119

R-Sq = 45.70%

Adj MS 4.992 2.802 67.771 1.050

F 4.75 2.67 64.53

P 0.032 0.026 0.000

R-Sq(adj) = 41.97%

The ANCOVA results indicate that project type has a significant effect on usefulness of the statistical techniques taught in the course as perceived at post-test time, after accounting for section and perceived usefulness at pre-test. Comparing the Regression and ANCOVA model results: The conclusion from ANCOVA matches the conclusion from regression analysis. Interestingly, the p-values are not identical. What is the difference between the two models? The answer is that the ANCOVA model accounts for “course section” by fitting 6 parallel lines, whereas the regression model allows us to drop some dummy variables and only fit 3 parallel lines, since sections 2, 4, 5, and 6, were found not to be significantly different from each other. FOR MORE ON ANOVA AND ANCOVA: BUSI 6480 (research design & ANOVA)

10

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Partial Least Squares (PLS) modeling Partial Least Squares Regression (PLS regression) is a multivariate statistical method that is related to the Structural Equation Modeling family of approaches, but PLS is component-based, instead of being covariance-based (Abdi 2010). PLS works well when you have more variables than observations, and when there is multicollinearity among the predictor variables (note that standard least squares regression will fail in both of these cases). Here we illustrate PLS Regression on the Course Acceptance Model data using smartPLS. In order for you to follow the illustration, you will need to download smartPLS from www.smartpls.de/forum. Register, follow the instructions, log on, and download the installation file. It may take you 1-2 days to obtain your activation key, but it is free of charge. Using smartPLS: 6.1. Open CAMdata01.xls, cleanup the variable names, data formatting, and remove empty columns. Also, make sure the first row of data does not have any missing values. Then convert the file into CAMdata.csv. Open smartPLS. Right-click on the Projects panel. Create a new project. On the content hierarchy list, right-click on project > Import indicator data. Select your CAMdata.csv file. Agree to the replacement of all missing values by –1.0. On the right panel, click Validate. If you do not get “The data file is valid”, repeat cleaning steps in Excel, reconvert the file into .csv format and re-open in smartPLS. Double-click the project name on the left panel, to open the workspace. Click on the button to switch to insertion mode. Use the mouse to click and drag oval-shaped construct. Click on the button to switch to selection mode. Right-click on the constructs > Rename to: Usefulness, Ease, Future Use, Attitude. Click on the button to draw the relationships. Drag-and-drop indicators from the left panel to the diagram. Select Calculate > PLS Algorithm > Finish. The fitted model shown below lists path coefficients, Cronbach’s alphas, R-squared and average variance explained (AVE) for each latent construct. Note that Ease of Use has 0% R-squared. This is normal and is due to the fact that Ease of Use is not explained by any other independent variable (i.e., it is a pure antecedent). The model explains 59.9% of the variability in Intention for Future Use. In order to see which paths are significant, you need to obtain t statistics. These can be obtained through bootstrapping, a resampling technique that samples (with replacement) multiple times, computing a new set of path coefficients each time. Select Calculate > Bootstrapping > BT Bootstrapping cases: Cases= 115 > Finish. The second diagram listed below shows the t values for each path coefficient. In order to decide which ones are significant, set them against a t distribution with n – p degrees of freedom. We see that all path coefficients are significant.

11

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

FOR MORE ON PLS: BUSI 6280 (causal & covariance structure modeling)

References Abdi, Hervé (2010). Partial least squares regression and projection on latent structure regression (PLS Regression). Wiley Interdisciplinary Reviews: Computational Statistics 2(jn/Feb), pp. 97–106. DOI: 10.1002/wics.51. Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Prentice-Hall, Englewood Cliffs, N.J. Cook, T.D. & Campbell, D.T. (1979). Quasi-experimentation design analysis issues for field settings. Houghton Mifflin Company, Boston. Davis, F.D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), pp. 319-339.

12

BUSI 6450/6220

CASE STUDY 1: INTRO TO SURVEY ANALYSIS

Evangelopoulos, N., Sidorova, A., and Riolli, L. (2003). Can Service-Learning Help Students Appreciate an Unpopular Course? A Theoretical Framework. Michigan Journal of Community Service Learning, 9(2), pp. 15-24. Gefen D., DW Straub, and MC Boudreau (2000). Structural equation modeling and regression: Guidelines for research practice, Communications of the Association of Information Systems, 4(7), pp. 1-76. Gefen, D., Rigdon, EE, and Straub, D. (2011). An update and extension to SEM guidelines for administrative and social science research. MIS Quarterly 35(2), pp. iii-xiv. Venkatesh, V. & Davis, F.D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2), pp. 186-205.

13