Business Market Research on Instant Messaging -2013
1 like545 views
The document discusses a social research study conducted to derive a model for measuring loyalty of IIT Masters and PhD students in using instant messaging on mobile. It aimed to verify if the model used in a Malaysian university study could be applied here. The researchers broke down loyalty into usefulness and satisfaction constructs, then further into variables like network size, perceived usefulness, and more. They tested the reliability and independence of questionnaire responses to validate the constructs. Factor and regression analyses showed satisfaction is defined by attention focus and perceived complementarity, while usefulness is defined by network size, perceived enjoyment and complementarity. The final standardized model showed loyalty is defined by satisfaction and usefulness.
Business Market Research on Instant Messaging -2013
- 1. 2013 Loyalty of Instant Messaging BUSINESS RESEARCH METHODS Group – 8 Subhasish Barman – 12BM60002 Arindam Routh – 12BM60011 Rajib Layek – 12BM60030 Debjyoti Mitra – 12BM60052 Manas Ghose – 12BM60100
- 2. Mobile IM Users (Post graduate Students) Intent: The intent of this social research is to derive a model for finding the loyalty quotient of the IIT Masters and PhD students in the use of instant messenger facility of the mobile and to verify whether the model used in Malaysian university can be applied here. Helpfulness: The research will help any mobile company to help launching newer version of instant messaging service in the IIT premises among masters and Ph .d students. Basic construct approach: We will be proceeding taking in consideration the basic construct of the Malaysian university and modify it as per the responses and local utilities. Our basic construct contains 4 primary construct so we can have a sample of 60. Construct detail: We will taking top-down approach. Loyalty of any service is based on its usefulness and satisfaction Usefulness Satisfaction Loyalty Attention Focus Referent Network Size Perceived Complementarit y Perceived Enjoyment
- 3. Analysis Usefulness and Satisfaction are highly dependent on similar quotient. So from this we need to define the basic construct which can independently describe Usefulness and Satisfaction but one can describe many concepts. Reliability and Validity of primary construct: The questionnaire prepared on each primary construct is as per the following link https://docs.google.com/forms/d/1106QAQJBHpllrfJgfnUvnUqJysmOyPoby-8uz- JC9U0/viewform Approach How we approach the problem can be depicted below No Yes No 1. Identifying the Primary Construct Stop 2. Cronbatch’s Alpha > .7 Wilk’s Lambda < .48 (Validity check between primary construct) It’s not a valid primary construct
- 4. Yes No Yes (Run Regression) No Yes No Yes Factor Loading > .5 (To check how many factors defined by primary construct) The primary construct is not defining secondary construct Choose other primary construct R2> .675 (To check if the factor are same as secondary construct) Assumption is wrong β> .11 (check independent variable can be kept or not ) Define Model The independent construct can be removed
- 5. 1. First Objective: Checking that the questionnaire prepared under each construct have responses which are defining the same thing and are correlated among themselves i.e. the intra construct items need to have higher correlation. For this purpose we will have done reliability test using the Cronbach’s alpha on the number of responses we have. For Cronbach’s Alpha>.7 implies our construct is reliable. In this case we are given 4 primary construct 1. Reference network size 2. Perceived Complementarity 3. Perceived Enjoyment 4. Attention Focus Items -Network1 -Network2 -Network3 We have done the reliability check and got the following Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .885 .888 3 This shows we have the Cronbach’s alpha>0.7 . Which implies my primary construct questionnaire is reliable. However if we look the correlation matrixNetwork2 and Network1 are more related. Inter-Item Correlation Matrix NETWORK1 NETWORK2 NETWORK3 NETWORK1 1.000 .781 .674
- 6. NETWORK2 .781 1.000 .722 NETWORK3 .674 .722 1.000 So if we remove Network3 we get below result Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .870 .877 2 But we are finding that the alpha is reduced hence we cannot deduct the Network3. By following the same way we can check for reliability of the each construct item and may realize ,which construct item to keep or not. 2. Second Objective: To check whether the questionnaire under each construct is unrelated and independent of the other primary construct questionnaire, i.e. inter construct questionnaire responses should be unrelated. Here we will be doing discriminant analysis on inter construct questionnaire and check for Wilks Lambda. If the value of lambda<.48 we will conclude that primary construct items are independent and are not overlapping. E.g. we chose two different construct items i.e. Perceived Complementarities (-PC1) and Perceived Enjoyment (PEU3). The analysis provides us the following statistic Wilks' Lambda Test of Function(s) Wilks' Lambda Chi-square df Sig. 1 .328 27.885 6 .000 Since here Wilks lambda <.48 we can confirm that the two items have no inter dependency and hence no inter construct dependency for this item We need to repeat the test for each construct and various combinations to check if our construct are valid or not.
- 7. 3. Third Objective: Since we have check the reliability of the items of the primary construct we can define a cumulative quotient defining the all items Responses weighted average is the construct value for that response. COMPUTE PEU=mean(PEU1,PEU2,PEU3). EXECUTE. COMPUTE NETWORK=mean(NETWORK1,NETWORK2.NETWORK3). >Error # 4285 in column 31. Text: NETWORK2.NETWORK3 >Incorrect variable name: either the name is more than 64 characters, or it is >not defined by a previous command. >This command not executed. EXECUTE. COMPUTE NETWORK=mean(NETWORK1,NETWORK2,NETWORK3). EXECUTE. COMPUTE PCOM=mean(PC1,PC2,PC3). EXECUTE. COMPUTE AF=mean(AF1,AF2,AF3,AF4). EXECUTE. So now we have variables 4 variables one variable per construct.
- 8. 4. Fourth objective: We will be trying to proof that our primary construct define the two secondary construct.We will be doing a factor analysis for the same and check whether the number of factors we got is two or more defining max variance. Secondly we will check whether our assumption on the secondary constructs dependency on primary construct hold good or not. Component Matrixa Component 1 2 PEU .874 NETWORK .888 PCOM .889 AF .845 .515 Extraction Method: Principal Component Analysis. Now the rotated component matrix is given below as Rotated Component Matrixa Component 1 2 PEU .835 .342 NETW ORK .866 .325 PCOM .718 .524 AF .369 .919 Extraction Method: Principal Component Analysis. Rotation Method: Varimax with Kaiser Normalization. a. Rotation converged in 3 iterations.
- 9. From the above analysis we got that we are defining two secondary construct but their dependency on the primary construct is redefined. Result Factor 1 is defined by-(PEU, NETWORK, PCOM) Factor 2 is defined by-(AF, PCOM) 5. Fifth Objective: We will be checking with the regression analysis to validate our analysis. We will cumulative quotient of Usefulness as PU and Satisfaction as SATIS Considering PEU, NETWORK, PCOM, AF as the independent variables we will run regression considering dependent variable as PU (Usefulness) and SATIS (satisfaction). From here we can say the model can define 90.4% of variability in PU (Usefulness) Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig.B Std. Error Beta 1 (Constant) -.015 .368 -.042 .967 PEU .167 .099 .169 1.696 .103 NETWORK .609 .100 .640 6.089 .000 PCOM .137 .112 .128 1.229 .231 Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .951 a .904 .889 .46323 .904 56.822 4 24 .000 a. Predictors: (Constant), AF, NETWORK, PEU, PCOM
- 10. AF .110 .095 .108 1.162 .256 a. Dependent Variable: PU From the beta value it is clear that the AF describe least variability in PU Hence if we remove AF and run the regression and we get the result as below From it is clear that R-square remains same hence AF is not defining the PU. Similarly for SATIS we get that Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .884a .781 .745 .70406 .781 21.412 4 24 .000 a. Predictors: (Constant), AF, NETWORK, PEU, PCOM Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig.B Std. Error Beta 1 (Constant) -.128 .560 -.228 .822 PEU .303 .150 .306 2.026 .054 Model Summary Mod el R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .948a .900 .888 .46207 .900 77.683 3 26 .000 a. Predictors: (Constant), PCOM, PEU, NETWORK
- 11. NETWOR K -.062 .152 -.065 -.408 .687 PCOM .359 .170 .334 2.114 .045 AF .427 .144 .416 2.967 .007 a. Dependent Variable: SATIS From here it is clear that AF,PCOM describes SATIS much more than other factors So removing other factors we get Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig.B Std. Error Beta 1 (Constant) .188 .557 .338 .738 AF .514 .139 .501 3.684 .001 PCOM .471 .146 .438 3.222 .003 a. Dependent Variable: SATIS Model Summary Model R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .862a .742 .722 .73411 .742 37.428 2 26 .000 a. Predictors: (Constant), PCOM, AF So we see that there is not much change in R squared value. So we can confirm that the satisfaction SATIS is defined by AF and PCOM. Hence our factor analysis is valid.
- 12. 6. Sixth Objective: Doing a factor analysis on Satisfaction and Usefulness we got the following Component Matrixa Component 1 2 SATIS1 .845 SATIS2 .948 SATIS3 .763 .591 PU1 .780 -.415 PU2 .877 PU3 .818 PU4 .871 Extraction Method: Principal Component Analysis. a. 2 components extracted. From the above it is clear that all the items are tending towards same concept. 7. Seventh Objective Running a regression on our assumption of defining LOYAL. On the basis of SATIS and PU we get the following result. Model Summary Mode l R R Square Adjusted R Square Std. Error of the Estimate Change Statistics R Square Change F Change df1 df2 Sig. F Change 1 .831a .690 .586 .86318 .690 6.672 7 21 .000 a. Predictors: (Constant), PU4, SATIS1, PU1, PU3, SATIS3, PU2, SATIS2
- 13. Coefficientsa Model Unstandardized Coefficients Standardized Coefficients t Sig.B Std. Error Beta 1 (Constant) .908 .664 1.366 .184 SATIS .618 .186 .642 3.319 .003 PU .151 .187 .156 .806 .427 a. Dependent Variable: LOYAL We have squared R >.6 which confirms our assumption but, there also remains a scope of improvement. So our Final standardized model is LOYAL=.642 SATIS+.156 PU
- 14. Also our final Construct Becomes Further Analysis of Cross tabulation on gender From the cross tabulation on the gender and the Final concept Loyalty. We get the following graph, from which we can conclude that the male are more loyal user of instant messaging than Female. However this conclusion can be redone with more number of samples Usefulness Referent N/w size Satisfaction Perceived compliment Perceived Enjoyment Attention Focus LOYALTY
- 15. LOYAL * GENDER Crosstabulation Count GENDER TotalFemale Male LOYAL 1 1 0 1 2 0 1 1 3 1 0 1 3.33333333333333 2 0 2 3.66666666666667 0 2 2 4 0 1 1 4.33333333333333 0 2 2 4.66666666666667 1 5 6 5 2 2 4 5.33333333333333 2 0 2 5.66666666666667 0 1 1 6 1 1 2 6.33333333333333 1 1 2 6.66666666666667 0 1 1
- 16. 7 0 1 1 Total 11 18 29 SUMMARY The main intention of this social research was to derive a model for finding the loyalty quotient of the IIT Masters and PhD students in the use of instant messenger facility and to verify whether the model used in Malaysian university can be applied here or not. The concept of loyalty was broken down to two constructs viz. Usefulness and Satisfaction. It was further broken down to variables like Referent network size, Perceived Usefulness, attention focus, perceived complimentarily. The approach has been to first identify a primary construct, pick up the responses assigned for that particular variable and subject them to Cronbach Alpha’s test. In statistics, Cronbach's is a coefficient of internal consistency. If the value obtained was greater than 0.7 then it was logically deduced that that the data obtained against that variable was reliable. The following chart has been used to determine the scales: Cronbach's alpha Internal consistency Alpha>= 0.9 Excellent 0.8 =<Alpha=< 0.9 Good 0.7 =<Alpha=< 0.8 Acceptable From the analysis data obtained it has been observed that PU, MEMBER, PC, TRUST, ENJOY, AF, SATIS (satisfaction), LOYAL, LEARN, WORK, REL (reliability)have a high internal consistency. The result is tabulated as follows: Please refer addendum. Cronbach's alpha Internal consistency TRUST, ENJOY,AF, SATIS Excellent PU, LEARN, REL Good MEMBER, PC, LOYAL, WORK Acceptable
- 17. Please refer addendum. Now, in order to establish the fact that the questionnaire under each construct are unrelated and independent of the other primary construct questionnaire, we conducted discriminant analysis on inter construct questionnaire and checked for Wilks Lambda. If the value of lambda< 0.38 we concluded that primary construct items are independent and are not overlapping. The following sets of data were organized for Wilks’ Lambda test: PEU3 and Network1 Network2 and PU1 Network2 and AF1 PEU2 and AF1 PU1 and AF4 PC1 and PEU3 All the tests with the above mentioned combinations returned the values greater than 0.38. Thus it could be concluded that the primary constructs are independent and not overlapping. Next, we will tried to prove that our primary construct defined the two secondary constructs. We conducted a factor analysis for the same and checked whether the number of factors we got was two or more by checking the maximum variance (R square > 0.68).Secondly, we checked whether our assumption on the secondary constructs depended on primary construct held good or not. From the rotated matrix results we obtained that factor 1 is defined by PEU, NETWORK and PCOM. Factor 2 is defined by AF and PCOM. Then regression analysis was carried out to determine the factors and map them to Usefulness and Satisfaction. It was found that satisfaction SATIS is defined by AF and PCOM and PU (Usefulness) is defined by NETWORK, PEU and PCOM. Thus, after further regression the following observation was made: So our Final standardized model is LOYAL=.642 SATIS+.156 PU We have squared R >.6 which confirms our assumption but, there also remains a scope of improvement by conducting regression analysis and finding out the R Square values for the remaining primary variables. A sample analysis has been done in the addendum section.
- 18. ADDENDUM Considering PU1,PU2,PU3, PU4 RELIABILITY /VARIABLES=PU1 PU2 PU3 PU4 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .898 .899 4 Inter-Item Correlation Matrix PU1 PU2 PU3 PU4 PU1 1.000 .711 .621 .681 PU2 .711 1.000 .761 .701 PU3 .621 .761 1.000 .659 PU4 .681 .701 .659 1.000 Considering MEMBER1,MEMBER2,MEMBER3 RELIABILITY /VARIABLES=MEMBER1 MEMBER2 MEMBER3 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR.
- 19. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .757 .765 3 Inter-Item Correlation Matrix MEMBER1 MEMBER2 MEMBER3 MEMBER1 1.000 .694 .581 MEMBER2 .694 1.000 .287 MEMBER3 .581 .287 1.000 Considering PC1,PC2,PC3 RELIABILITY /VARIABLES=PC1 PC2 PC3 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .798 .800 3 Inter-Item Correlation Matrix PC1 PC2 PC3 PC1 1.000 .763 .493 PC2 .763 1.000 .457
- 20. Inter-Item Correlation Matrix PC1 PC2 PC3 PC1 1.000 .763 .493 PC2 .763 1.000 .457 PC3 .493 .457 1.000 PC3 if we remove, Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .866 .868 2 Inter-Item Correlation Matrix PC1 PC2 PC1 1.000 .766 PC2 .766 1.000 Considering TRUST1,TRUST2,TRUST3,TRUST4 RELIABILITY /VARIABLES=TRUST1 TRUST2 TRUST3 TRUST4 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics
- 21. Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .944 .945 4 Inter-Item Correlation Matrix TRUST1 TRUST2 TRUST3 TRUST4 TRUST1 1.000 .771 .841 .754 TRUST2 .771 1.000 .834 .856 TRUST3 .841 .834 1.000 .810 TRUST4 .754 .856 .810 1.000 Consideing ENJOY1,ENJOY2,ENJOY3,ENJOY4 RELIABILITY /VARIABLES=ENJOY1 ENJOY2 ENJOY3 ENJOY4 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .951 .952 4 Inter-Item Correlation Matrix ENJOY1 ENJOY2 ENJOY3 ENJOY4
- 22. ENJOY1 1.000 .860 .943 .776 ENJOY2 .860 1.000 .804 .796 ENJOY3 .943 .804 1.000 .814 ENJOY4 .776 .796 .814 1.000 Considering AF1,AF2,AF3,AF4 RELIABILITY /VARIABLES=AF1 AF2 AF3 AF4 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .923 .923 4 Inter-Item Correlation Matrix AF1 AF2 AF3 AF4 AF1 1.000 .875 .719 .781 AF2 .875 1.000 .769 .729 AF3 .719 .769 1.000 .624 AF4 .781 .729 .624 1.000 Considering SATIS1,SATIS2,SATIS3 RELIABILITY
- 23. /VARIABLES=SATIS1 SATIS2 SATIS3 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .903 .910 3 Inter-Item Correlation Matrix SATIS1 SATIS2 SATIS3 SATIS1 1.000 .812 .696 SATIS2 .812 1.000 .808 SATIS3 .696 .808 1.000 Considering LOYAL1,LOYAL2,LOYAL3 RELIABILITY /VARIABLES=LOYAL1 LOYAL2 LOYAL3 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .792 .797 3
- 24. Inter-Item Correlation Matrix LOYAL1 LOYAL2 LOYAL3 LOYAL1 1.000 .788 .431 LOYAL2 .788 1.000 .480 LOYAL3 .431 .480 1.000 We can remove LOYAL3 Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .880 .882 2 Considering LEARN1,LEARN2,LEARN3 RELIABILITY /VARIABLES=LEARN1 LEARN2 LEARN3 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .867 .869 3
- 25. Inter-Item Correlation Matrix LEARN1 LEARN2 LEARN3 LEARN1 1.000 .639 .630 LEARN2 .639 1.000 .798 LEARN3 .630 .798 1.000 Considering WORK1,WORK2,WORK3,WORK4 RELIABILITY /VARIABLES=WORK1 WORK2 WORK3 WORK4 WORK5 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .667 .667 5 We reject this as Cronbach’s Alpha is less that 0.7 Considering REL1,REL2,REL3,REL4,REL5 RELIABILITY /VARIABLES=REL1 REL2 REL3 REL4 REL5 /SCALE('ALL VARIABLES') ALL /MODEL=ALPHA /STATISTICS=DESCRIPTIVE SCALE CORR ANOVA FRIEDMAN /SUMMARY=MEANS COV CORR.
- 26. Reliability Statistics Cronbach's Alpha Cronbach's Alpha Based on Standardized Items N of Items .806 .808 5 Inter-Item Correlation Matrix REL1 REL2 REL3 REL4 REL5 REL1 1.000 .474 .596 .289 .443 REL2 .474 1.000 .523 .438 .555 REL3 .596 .523 1.000 .167 .385 REL4 .289 .438 .167 1.000 .707 REL5 .443 .555 .385 .707 1.000 Validity in between PEU3 and Network1 Wilks' Lambda Test of Functio n(s) Wilks' Lambda Chi-square df Sig. 1 .461 19.357 6 .004 In between PU1 and AF4
- 27. Wilks' Lambda Test of Functio n(s) Wilks' Lambda Chi-square df Sig. 1 .434 20.011 6 .003 In Between PEU2 and AF1 Eigenvalues Functio n Eigenvalue % of Variance Cumulative % Canonical Correlation 1 .479a 100.0 100.0 .569 In Between Network2 and AF1 Wilks' Lambda Test of Functio n(s) Wilks' Lambda Chi-square df Sig. 1 .419 19.995 6 .003 In Between Network2 and PU1 Wilks' Lambda Test of Functio n(s) Wilks' Lambda Chi-square df Sig.