Structural Equation Modelling and Logistic Regression - Report Example

Factors that influence attitude towards replacing cash-Logistic regression Assem Almnashi Advanced Quantitative 01/06 Introduction The current study majorly focused on the factors that influence attitude towards replacing cash and it was measured using a single question with a yes / no answer. In this case, the response or variable was binary (0/1).The predictor variables that were of interest include perceived usefulness and perceived ease of use cashless payments as well as trust. The perceived usefulness and perceived ease of use cashless payments as well as trust as predictor variables were measured with the use of 4 items scales and 5 item scale respectively where, their psychometric characteristics were assessed before being included in the logit by averaging them since they were multi-item scale variables. Age gender and education level were measured by use of dummy variables and hence, they represented demographic information. This study is a representation of the relevance of the technology acceptance model with knowledge based trust integration in individual behavior explanation. Hence, this provides the support for the association added in representing the effect of trust ease of use and usefulness on attitude in the cashless economy acceptance. The technology acceptance model constructs, is considered to be the key criterion needed in determining the behavioral intention of adopting the cashless transaction. Other conciliators like attitudes seen to contribute equally on influencing the behavioral intension in adopting cashless transaction. This clearly means that to appropriately influence the users to adopt cashless economy, there is much attention required to be paid in such aspects. Trust as well is seen to contribute some degree of significant impact on the attitude of the users towards cashless economy acceptance. Technology acceptance model is known to have been developed by Davis (1989) and it is considered to be the research model that is most popular in predicting acceptance and use of technology as well as information system by individual users. The technology acceptance model has been broadly verified as well as studied by various studies which investigated the technology acceptance behavior of an individual in various information systems constructs. Two factors do exist in technology acceptance model and include perceived ease of use and perceived usefulness where perceived ease of use is considered to be critical in the behaviors of computer use. Perceived usefulness was defined by Devis as the subjective probability of the prospective user that applying a particular supplication system will lead to enhancement of the life or job performance. The perceived ease of use by definition is the level to which it is expected by the prospective user that the target system should be free from the effort. Basing on the technology acceptance model perceived usefulness as well as ease of use is seen to be the most critical determinants of the actual system application and both of these variables are affected by the external variables. The key external factors which influence them are political, social and cultural factors. Social factors are such as skills and language while the political factors include majorly the effects of applying technology in political crisis as well as politics. The attitude to use is based on the desirability evaluation of the user of applying a given information system. Basing on Inglehart and Baker (2000), trust can be defined as a social construct centered on the global relationships of the individual groups. It is considered to be an abstract concept as well as complex and consensus on its definition is not there also on how it influences behaviors and the way it is formed (Bahmanziari, Odom and Ugrin, 2009). Thus, trust emerges is cases of vulnerability and it entails the urge of risk taking (Kenning, 2008).Trust is considered to be in the risk context and it facilitates the peoples living in a given situation (Deutsch 1962; Mayer et al. 1995). Trust can be considered to be a social construct brand which facilitates cooperation and coordination among individuals. Test of psychometric properties of perceived usefulness Table 1. Reliability Statistics Cronbachs Alpha Cronbachs Alpha Based on Standardized Items N of Items .829 .832 4 Table 1 above represents the reliability statistics table and it gives the Cronbach’s alpha actual values. In this case, it is observed that Cronbach’s alpha is 0.829 and this shows a high level of internal consistency for the given scale (perceived usefulness). Table 2. Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Squared Multiple Correlation Cronbachs Alpha if Item Deleted Usefulness_1 12.83 4.302 .607 .383 .810 Usefulness_2 13.01 3.348 .761 .580 .733 Usefulness_3 13.08 3.750 .645 .434 .789 Usefulness_4 13.17 3.436 .640 .427 .797 Table 2 above represents the “Cronbach’s Alpha if item deleted” as shown in the last column. The values shown in the column represents Cronbach’s alpha in case a given item is deleted. The table 2 shows that when any item or question is removed there might be a lower Cronbach’s alpha and this implies that no item should be omitted. Test of psychometric properties of perceived ease of use Table 3. Reliability Statistics Cronbachs Alpha Cronbachs Alpha Based on Standardized Items N of Items .843 .854 4 Table 3 above represents the reliability statistics table and it gives the Cronbach’s alpha actual values. In this case, Cronbach’s alpha is seen to be 0.843 and hence, this means that there is a high level of internal consistency for the given scale (perceived ease of use). Table 4. Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Squared Multiple Correlation Cronbachs Alpha if Item Deleted Ease_of_Use_1 12.12 4.397 .659 .442 .810 Ease_of_Use_2 12.25 4.044 .762 .653 .766 Ease_of_Use_3 12.38 4.116 .746 .636 .774 Ease_of_Use_4 12.62 3.710 .597 .357 .856 Table 4 represents the “Cronbach’s Alpha if item deleted” as indicated in the last column. It is shown that when any item or question is removed there might be a lower Cronbach’s alpha and this means that no item should be omitted. Test of psychometric properties of trust scale Table 5. Reliability Statistics Cronbachs Alpha Cronbachs Alpha Based on Standardized Items N of Items .881 .885 5 Table 5 above represents the reliability statistics table and it gives the Cronbach’s alpha actual values. it is observed that Cronbach’s alpha is 0.881 and this shows a high level of internal consistency for the given scale (Trust scale). Table 6. Item-Total Statistics Scale Mean if Item Deleted Scale Variance if Item Deleted Corrected Item-Total Correlation Squared Multiple Correlation Cronbachs Alpha if Item Deleted Trust_1 14.58 9.907 .624 .432 .881 Trust_2 14.13 10.509 .715 .518 .857 Trust_3 14.53 9.466 .770 .594 .842 Trust_4 14.54 9.641 .773 .671 .842 Trust_5 14.21 10.586 .724 .618 .855 Table 6 represents the “Cronbach’s Alpha if item deleted” as indicated in the last column. It is observed that in case any item is removed there could be a lower Cronbach’s alpha and this means that no item should be omitted. Logistic regression Logistic regression is employed in predicting a categorical variable from the variables which are predictors. With dependent variables that are dependent, the analysis of the discrimination function is normally applied in case all the predictor variables are continuous as well as distributed appropriately; it is as well applied in case all the predictor variables are categorical and the selection of the logistic regression is based on the predictor variables if they are both categorical as well as continuous. Table 7. Case Processing Summary Unweighted Casesa N Percent Selected Cases Included in Analysis 320 99.7 Missing Cases 1 .3 Total 321 100.0 Unselected Cases 0 .0 Total 321 100.0 a. If weight is in effect, see classification table for the total number of cases. The table 7 above represents a breakdown of cases applied as well as not applied in the analysis. It is observed that all the case (321) was included in the analysis and there were zero missing cases. The table 8 below indicates the coding for the outcome variable that include yes/no. Table 8. Dependent Variable Encoding Original Value Internal Value Disagree 0 Agree that a cashless society is the right way to go 1 Table 9. Categorical Variables Codings Frequency Parameter coding (1) (2) (3) (4) education 1.00 3 1.000 .000 .000 .000 2.00 42 .000 1.000 .000 .000 3.00 122 .000 .000 1.000 .000 4.00 106 .000 .000 .000 1.000 5.00 47 .000 .000 .000 .000 age_group 18-29 113 1.000 .000 .000 30-49 107 .000 1.000 .000 50-59 100 .000 .000 1.000 gender male 164 1.000 female 156 .000 The above table represents the way the categorical variable values ranks were considered. In the model, we have terms or dummy variables for education, age group and gender that include 1, 2,3,4,5 where rank 5 is the omitted category. Block 0: Beginning Block (First model). In the output (represented below), the first model represents a null hypothesis and in other terms, it is a model without predictors. Table 11 below indicates the attitude’s unconditional log odds. It is shown in table 11 that the cut value is 0.5,and this can be interpreted as incase there is a probability of classifying a case into a the category of yes as greater than 0.5,then that given case will be categorized into the category of yes. If this is not the case, then the case will be classified in the category of no as indicated previously. In the labeled table variables not in the equation provides the score test results and it is as well referred to as lagrange multiplier test. In this case, in the score column, the estimated change in the model fit in case there is addition of the term to the model is provided. Also the degree of freedom as well as the p value has been indicated for the estimated change. According to the below output, it is observed that all the predictors have the capacity of improving the fit of the model since the p value is less than 0.05. Table 10. Classification Tablea,b Observed Predicted attitude Percentage Correct Disagree Agree that a cashless society is the right way to go Step 0 attitude Disagree 0 136 .0 Agree that a cashless society is the right way to go 0 184 100.0 Overall Percentage 57.5 a. Constant is included in the model. b. The cut value is .500 Table 11. Variables in the Equation B S.E. Wald df Sig. Exp(B) Step 0 Constant .302 .113 7.145 1 .008 1.353 Table 12. Variables not in the Equation Score df Sig. Step 0 Variables usefulnessaverage 103.457 1 .000 eraseofuse 39.060 1 .000 trust 35.997 1 .000 gender(1) .025 1 .874 age_group 8.090 3 .044 age_group(1) .885 1 .347 age_group(2) 6.304 1 .012 age_group(3) .283 1 .595 education 8.619 4 .071 education(1) .724 1 .395 education(2) 7.449 1 .006 education(3) .804 1 .370 education(4) .243 1 .622 Overall Statistics 115.050 11 .000 Block 1: Method = Enter (model with predictors) Table 13. Omnibus Tests of Model Coefficients Chi-square df Sig. Step 1 Step 136.683 11 .000 Block 136.683 11 .000 Model 136.683 11 .000 The table above represents the overall test for the model with predictors, and it is observed that the chi-square value (136.68) that has a p value of 0.00 seen to be less than 0.05 implies that the model fits significantly better compared to the empty model or the model that has no predictors. In the summary model below with -2 log likelihood of 299.704 could employed in the comparisons of the nested models. The model summary as well provides the pseudo R-square measures. The model summary below consists of the Nagelkerke R square and Cox & Snell R square values where, they are both used in the calculation of the explained variations and both values are also known as pseudo R squared values. The model ranges from 34.8% to 46.7% and both values represent Cox & Snell R squared and Nagelkerke R squared modification respectively. The Nagelkerke R squared value is interpreted basing on the fact that Cox & Snell R squared can’t reach the value of one and therefore, Nagelkerke R squared is considered where R squared of the model is taken to be 46.7%. Model Summary Step -2 Log likelihood Cox & Snell R Square Nagelkerke R Square 1 299.704a .348 .467 a. Estimation terminated at iteration number 5 because parameter estimates changed by less than .001. The Hosmer-Lemeshow test usually used in testing the null hypothesis that the predictions of the model will fit perfectly with the group memberships observed. In this case, there is arrangement of the cases in order based on the probability predicted on the variable criterion. There is then a division of these cases into equal groups based on the predicted targeted event probability. The actual group membership as well as predicted group a membership is obtained in this case. The results output is 2x10 contingency table as indicated below. There is computation of the chi-square statistics where, there is comparison of the observed frequencies with those frequencies that are expected from the linear model. A chi-square that is not significant shows that the data fit the model well. Based on this, the results as indicated below shows that the chi-square is not significant and this means that the data fits the model well. This procedure is known to suffer from various problems and one being that it is based on a significant test. The test could be significant in case of large sample size despite having good fit. The small sample size could lead to it not being significant despite it having poor fit. Hosmer and Lemeshos never recommend its application. Hosmer and Lemeshow Test Step Chi-square df Sig. 1 3.109 8 .927 Contingency Table for Hosmer and Lemeshow Test attitude = Disagree attitude = Agree that a cashless society is the right way to go Total Observed Expected Observed Expected Step 1 1 31 30.231 1 1.769 32 2 26 26.408 6 5.592 32 3 23 22.408 9 9.592 32 4 17 18.408 15 13.592 32 5 11 13.272 21 18.728 32 6 12 9.065 20 22.935 32 7 5 6.174 27 25.826 32 8 5 4.610 27 27.390 32 9 4 3.429 28 28.571 32 10 2 1.995 30 30.005 32 The variables in the equation table below indicate the standard errors, coefficients, the p values and the Wald test statistics as well as the odds ratio. It is observed that usefulness and trust as the predictor variables are statistically significant since the p value is less than 0.05 while the ease of use is not statistically significant. The logistic regression coefficient usually provides the change in the outcome odds ratio for a unit increase in the explanatory variables. Basing on the results, it is observed that for every increase in usefulness, the attitude (yes verses no) odds ratio increases by 2.54. For every increase in ease of use, the attitude (yes verses no) odds ratio decreases by 0.308 and for every increase in trust, the attitude (yes verses no) odds ratio increases by 0.532 Classification Tablea Observed Predicted attitude Percentage Correct Disagree Agree that a cashless society is the right way to go Step 1 attitude Disagree 97 39 71.3 Agree that a cashless society is the right way to go 31 153 83.2 Overall Percentage 78.1 a. The cut value is .500 Variables in the Equation B S.E. Wald df Sig. Exp(B) 95% C.I.for EXP(B) Lower Upper Step 1a usefulnessaverage 2.540 .365 48.391 1 .000 12.676 6.197 25.927 eraseofuse -.308 .302 1.036 1 .309 .735 .406 1.330 trust .532 .221 5.815 1 .016 1.702 1.105 2.623 gender(1) .084 .296 .081 1 .776 1.088 .609 1.943 age_group 4.678 3 .197 age_group(1) .429 .574 .558 1 .455 1.535 .498 4.732 age_group(2) 1.038 .580 3.199 1 .074 2.823 .905 8.806 age_group(3) .537 .595 .816 1 .366 1.711 .533 5.491 education 7.325 4 .120 education(1) -3.292 1.493 4.858 1 .028 .037 .002 .694 education(2) -1.057 .555 3.633 1 .057 .347 .117 1.030 education(3) -.479 .464 1.063 1 .302 .619 .249 1.539 education(4) -.606 .467 1.687 1 .194 .545 .219 1.361 Constant -11.439 1.504 57.804 1 .000 .000 a. Variable(s) entered on step 1: usefulnessaverage, eraseofuse, trust, gender, age_group, education. The above table also shows the Wald test and it is employed in determining statistically significance of every predictor variables. Basing on the results, it is observed that usefulness (0.00) and trust (0.016) added to the model significantly, but ease of use (0.309) never added significantly to the model Cook’s Statistics (residual statistics) Influential observations or points are considered to be points lying far from other data on the given graph in a horizontal direction. Such points are termed as outliers and their residual values are usually larger. The points have a significant influence on the regression line slope and these results to a poor fit. After running cook’s test, it was observed that there are no influential points observed. In this case, the influential point impact is not exhibited. The influential points were determined quantitatively by Cook’s statistics and Cook’s D is considered to be significant if the value significantly differs from other Cook’s D statistics. If DFITS is greater than 2 then the point will be considered to be an influential point. For our case, all the values were less than 2 and this clearly indicated that there were no influential points in the data set. Conclusion The logistic regression model was seen to be statistically significant, and the model explained a 46.7% (Nagelkertke R squared) of the variance in the response. For every increase in usefulness, the attitude (yes verses no) odds ratio increases by 2.54, for every increase in ease of use, the attitude (yes verses no) odds ratio decreases by 0.308 and for every increase in trust, the attitude (yes verses no) odds ratio increases by 0.532.Based on the results, the first model represents a null hypothesis and in other terms, it is a model without predictors. In the table which is labeled variables in the equation provides the attitude’s unconditional odds ratio. As it can be noticed in the table the cut value is 0.5,it implies that if there is a probability of classifying a case into the category of yes as greater than 0.5, then that given case will be categorized into the category of yes. If this is not the case, then the case will be classified in the category of no as indicated. This therefore means that the 50% cut off is considered to be appropriate The Hosmer-Lemeshow test is usually used in testing the null hypothesis that the predictions of the model will fit perfectly with the group memberships observed. A chi-square that is not significant shows that the data fit the model well. Based on this, the results show that the chi-square is not significant and this means that the data fits the model well. This procedure is known to suffer from various problems and one being that it is based on a significant test. The test could be significant in case of large sample size despite having good fit. The small sample size could lead to it not being significant despite it having poor fit. Hosmer and Lemeshos never recommend its application. Basing on the technology acceptance model perceived usefulness as well as ease of use is seen to be the most critical determinants of the actual system application and both of these variables are affected by the external variables. The key external factors which influence them are political, social and cultural factors. Social factors are such as skills and language while the political factors include majorly the effects of applying technology in political crisis as well as politics. The attitude to use is based on the desirability evaluation of the user of applying a given information system. The technology acceptance model constructs is considered to be the key criterion needed in determining the behavioral intention of adopting the cashless transaction. Other conciliators like attitudes seen to contribute equally on influencing the behavioral intension in adopting cashless transaction. This clearly means that to appropriately influence the users to adopt cashless economy, there is much attention required to be paid in such aspects. Trust as well is seen to contribute some degree of significant impact on the attitude of the users towards cashless economy acceptance. Because all the construct of perceived risk that include finance, performance risk and security risk is considered to be negative factors in the intensions of adopting cashless economy and trust helps in reducing fears as well as enabling persons to live in uncertain and risky situations. This study is a representation of the relevance of the technology acceptance model with knowledge based trust integration in individual behavior explanation. Hence, this provides the support for the association added in representing the effect of trust ease of use and usefulness on attitude in the cashless economy acceptance. Reference Bahmanziari T, Odom M.D, Ugrin J.C. (2009). An experimental evaluation of the effects of internal and external e-Assurance on initial trust formation in B2C ecommerce. International Journal of Accounting Information Systems,Vol. 10,pp. 152–170 Davis, F. D. (1989). “Perceived Usefulness, Perceived Ease of Use, and User Acceptance of InformationTechnology,” MIS Quarterly (13:3), , pp. 319‐339 Deutsch, M., 1962. Cooperation and trust: some theoretical notes. Nebraska Symposium on Motivation 10, 275–318 Inglehart R, and Baker W. (2000).Modernization, cultural change and the persistence of traditional values. American Sociological Review,Vol. 65,pp. 19–51 Kenning P. (2008). The influence of general trust and specific trust on buying behavior. International Journal of Retail & Distribution Management,Vol. 36, No. 6, pp. 461–476 Mayer, R.C. Davis, J.H. Schoorman, F.D. 1995 ‘An integrative model of organizational Trust’, Academy of Management Review, Vol 20, No. 3. pp. 709–734 Appendix 1 Item Statistics Mean Std. Deviation N Usefulness_1 4.53 .618 320 Usefulness_2 4.35 .805 320 Usefulness_3 4.29 .763 320 Usefulness_4 4.20 .865 320 Inter-Item Correlation Matrix Usefulness_1 Usefulness_2 Usefulness_3 Usefulness_4 Usefulness_1 1.000 .591 .488 .474 Usefulness_2 .591 1.000 .632 .630 Usefulness_3 .488 .632 1.000 .508 Usefulness_4 .474 .630 .508 1.000 Scale Statistics Mean Variance Std. Deviation N of Items 17.36 6.238 2.498 4 Item Statistics Mean Std. Deviation N Ease_of_Use_1 4.34 .712 320 Ease_of_Use_2 4.20 .742 320 Ease_of_Use_3 4.08 .732 320 Ease_of_Use_4 3.83 .967 320 Inter-Item Correlation Matrix Ease_of_Use_1 Ease_of_Use_2 Ease_of_Use_3 Ease_of_Use_4 Ease_of_Use_1 1.000 .617 .593 .505 Ease_of_Use_2 .617 1.000 .778 .541 Ease_of_Use_3 .593 .778 1.000 .532 Ease_of_Use_4 .505 .541 .532 1.000 Scale Statistics Mean Variance Std. Deviation N of Items 16.45 6.869 2.621 4 Item Statistics Mean Std. Deviation N Trust_1 3.42 1.062 320 Trust_2 3.87 .855 320 Trust_3 3.46 1.001 320 Trust_4 3.46 .966 320 Trust_5 3.79 .833 320 Inter-Item Correlation Matrix Trust_1 Trust_2 Trust_3 Trust_4 Trust_5 Trust_1 1.000 .584 .597 .504 .456 Trust_2 .584 1.000 .619 .633 .563 Trust_3 .597 .619 1.000 .684 .652 Trust_4 .504 .633 .684 1.000 .764 Trust_5 .456 .563 .652 .764 1.000 Scale Statistics Mean Variance Std. Deviation N of Items 18.00 15.207 3.900 5 Casewise Listb Case Selected Statusa Observed Predicted Predicted Group Temporary Variable attitude Resid ZResid 6 S D** .900 A -.900 -2.993 122 S D** .861 A -.861 -2.491 123 S D** .892 A -.892 -2.867 138 S A** .124 D .876 2.655 158 S D** .911 A -.911 -3.202 208 S D** .896 A -.896 -2.928 212 S D** .852 A -.852 -2.397 269 S A** .088 D .912 3.210 274 S D** .940 A -.940 -3.961 299 S D** .865 A -.865 -2.528 303 S D** .930 A -.930 -3.654 a. S = Selected, U = Unselected cases, and ** = Misclassified cases. b. Cases with studentized residuals greater than 2.000 are listed. Step number: 1 Observed Groups and Predicted Probabilities 16 + + I I I I F I A I R 12 + A A + E I A A I Q I AA A A I U I AAA A A I E 8 + AAA A A A + N I A AA AAA A AA A I C I D A A A AA AAA A AAAA I Y ID D A D A D AA A A A A AAAA AAA A AAAA I 4 +D DD D D D D A A D AA A A A A A A A AAAAAA AAAAAAAAAAA + ID DD D D D D DAA DDAAA A D AADAAD A AA A A AA AAA A A A AAAAAAAAADAAADAAAAAA I ID DDD D D DDDD D DDD ADDDDDDD D DADDAD DADD A DAA A A DA DDD DA DAAAADAAAAAAADDAADAAAAAA I IDDDDDDDDDDDDDDDD DDDDADDDDDDDDD DDADDDDDDADDDD AD DDAAD D ADDDADDDADAAADADDDDADDDADADDAADADADDA A I Predicted ---------+---------+---------+---------+---------+---------+---------+---------+---------+---------- Prob: 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1 Group: DDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA Predicted Probability is of Membership for Agree that a cashless society is the right way to go The Cut Value is .50 Symbols: D - Disagree A - Agree that a cashless society is the right way to go Each Symbol Represents 1 Case. Read More