Non Response Bias Test Business Essay
This chapter intends to indicate the results based on the analysis of data collected using Structural Equation Modeling. This data collected is screened, validated and analyzed by using factor analysis and the internal consistency procedure of reliability analysis, correlation estimation, convergent validity analysis, discriminant validity, confirmatory factor analysis (CFA), and Structural Equation Modeling (SEM). The results of the analysis are presented using statistical packages, Analysis of Moment Structures (AMOS) and Statistical Package for the Social Sciences (SPSS). Hypothesis testing is used to examine the relationship between the variables of learning organization, organizational innovativeness, and organizational performance.
4.2 Non-Response Bias Test
Non-response bias test was undertaken because this type of bias affects the interpretation of the variables and subsequently affects the overall conclusions resulted from the data analysis. Evidence from existing literatures have established that the non-respondents sometimes differ systematically from the respondents both in attitudes, behaviors, personalities, motivations, demographics and/or psychographics, in which any or all of which might affect the results of the study (Malhotra, Hall, Shaw, Oppenheim, 2006).
Non-response bias was assessed by comparing the responses in the questionnaires between the early and late returns (Armstrong & Overton, 1977, Churchill & Brown, 2004, Malhortra et al., 2006). Non-response bias has been tested using the t-test to compare the similarities between mean, standard deviation, and standard error mean of the demographic data of the last 60 percent of the respondents (number of cases = 238) to the data of the first 40 percent of the respondents (number of cases = 160). According to the results as shown in Table 4.1, the responses indicate no significant differences between each group. Therefore, non-response bias has no significant impact in this study.
Table 4.1
Test of Non-Response Bias between Group Differences of Early 40 percent and Late 60 percent
Variable
Response
Number of Cases
Mean
Standard Deviation
Standard Error Mean
Gender
Early
160
1.280
0.451
0.036
Late
238
1.300
0.458
0.030
Type of Business
Early
160
1.430
0.497
0.039
Late
238
1.470
0.500
0.032
Experience
Early
160
3.140
0.759
0.060
Late
238
2.990
0.787
0.051
Position
Early
160
1.310
0.465
0.037
Late
238
1.300
0.460
0.030
Number of Employees
Early
160
1.900
0.301
0.024
Late
238
1.890
0.307
0.020
Age of Business
Early
160
2.530
0.501
0.040
Late
238
2.550
0.499
0.032
International Business
Early
160
1.510
0.502
0.040
Late
238
1.450
0.498
0.032
4.3 Data Screening and Outliers
This section considers each part of the data screening and investigates the root cause of data error that may impact the final outcome. The data screening observes the data position for all remaining data. All data is entered into SPSS version 20 and analyzed using AMOS version 18.0. Data screening includes outlier checking, missing data, descriptive statistic, univariate normality, multicollinearity, reliability, and validity testing.
4.3.1 Outlier Checking (Mahalanobis Distance)
Statistical evidence has established outliers as any observations, which are numerically distant if compared to the rest of the data set (Bryne, 2010). There is existing research literature that uses different methods of detecting outliers among which are includes classifying data points based on an observed (Mahalanobis) distance from the research expected values (Hair et al., 2010; Hau & Marsh, 2004). The constructive argument in favor of outlier treatments based on Mahalanobis distance is that it serves as an effective means of detecting outliers through the settings of some predetermined threshold that will assist in defining whether a point could be categorized as outlier or not.
For this research, chi-square statistics has been used as the threshold value to determine the empirical optimal values in the research. This decision is in line with the arguments of Hair et al. (2010) which emphasized on the need to create a new variable in the SPSS excel to be labelled ‘No.’ numbering from the beginning to the end of all variables. The Mahalanobis can simply be achieved by running a simple linear regression through the selection of the newly created response number as the dependent variable and selecting all measurement items apart from the demographic variables as independent variables. Consequently, a new output is created called Mahalanobis1 (Mah1), upon which a comparison was made between the chi-square value and a maximum of Mahalanobis distance value in Table 4.2, which also includes the new Mahalanobis output.
Table 4.2
Outlier Detection (Mahalanobis Distance)
Min.
Max.
Mean
Standard Deviation
Number of Cases
Predicted Value
-6.32
451.63
199.50
77.192
398
Standard Predicted Value
-2.666
3.266
.000
1.000
398
Standard Error of Predicted Value
5.899
59.259
28.507
6.600
398
Adjusted Predicted Value
-31.67
463.81
198.32
78.266
398
Residual
-253.756
193.111
.000
85.293
398
Standard Residual
-2.817
2.144
.000
.947
398
Deleted Residual
-270.853
226.881
1.179
94.793
398
Mahal. Distance
.706
170.845
40.897
18.576
398
Note. Min. = minimum; Max. = maximum
The value of Mahalanobis Distance (D2) is greater than a critical value and used as the threshold level for D2/df measure which should be conservative level of significance (0.005 or 0.001) for designation on outliers (Hair et al., 2010). D2 is known as the Mahalanobis distance and is the distance between each observation in multidimensional space from the mean centre of all the observations. Df is the degree of freedom, or the number of variables involved (Hair et al., 2010).
For this study, the maximum of Mahalanobis distance is 170.845 as shown in Table 4.2 greater than the critical value. The critical value mentioned in chi-square value is 74.745 at df=41, p=0.001 (refer to Appendix D). This means there are 6 respondents of the total of 398 respondents that were deleted. The screening data of these outliers and the final regression in this study were performed using the remaining 392 samples in the data set.
4.3.2 Missing Data
Once the questionnaire data are collected, the first step in data screening is to identify the data errors. The extent of missing data affects the unit data putting at risk the analysis result. Normally, missing data under 10 percent for an individual case or observation can generally be ignored, except when the missing data occurs in a specific nonrandom fashion such as concentration in a specific set of questions and attrition at the end of the questionnaire (Hair et al., 2010). In this study, the missing data (refer to Appendix E) does not exist in every questionnaire. Therefore, this study determines the number of cases without missing variables, which provide the sample size variable for data analysis still remedies.
4.3.3 Descriptive Statistic
The following profile was found among the data screening process. In general, the descriptive latent constructs include maximum, minimum, mean, standard deviation, mode, and median. The nine latent constructs (continuous learning (CL), inquiry and dialogue (ID), team learning (TL), embedded system (ES), empowerment (EM), system connection (SC), strategic leadership (SL), organizational innovativeness (OI), and organizational performance (OP) are presented in Table 4.3.
Table 4.3
Descriptive Statistics of Variables
Number of Cases
Missing Data
Mean
Standard Deviation
Continuous Learning
392
3.169
1.263
Inquiry and Dialogue
392
3.379
1.147
Team Learning
392
3.304
1.223
Embedded system
392
3.401
1.134
Empowerment
392
3.480
1.129
System Connection
392
3.098
1.294
Strategic Leadership
392
3.361
1.174
Organizational Innovativeness
392
3.342
0.945
Organizational Performance
392
2.998
1.037
The mean value of the nine constructs are based on 41 questions. Organizational performance (OP) is lowest for mean value (2.998) while the highest mean is empowerment (EM = 3.480). For standard deviation, system connection (SC) is the highest value (1.294), continuous learning (CL) is 1.263, team learning (TL) is 1.223, strategic leadership (SL) is 1.174, inquiry and dialogue (ID) is 1.147, embedded system (ES) is 1.134, empowerment (EM) is 1.129, and organizational performance (OP) is 1.037, but the lowest value is the organizational innovativeness (OI) with 0.945.
4.3.3.1 Demographic Profile of the Respondents
The demographic profile of the participants SMEs characteristic consists of the seven major items as follows: (1) gender, (2) type of business, (3) experience, (4) position, (5) number of employees, (6) age of business, and (7) international business. The analysis results of the before mentioned variables are presented in Table 4.4 and Appendix F. The results are shown as the frequency and percentage.
Table 4.4
The Demographic Profile of the Respondents consist of the Seven Major Items
Demographics
Frequency
Percentage
Gender
Male
Female
280
112
71.4
28.6
Type of business
Manufacturing
Service
216
176
55.1
44.9
Experience of work (in this business)
More than 5 but less than 7 years
More than 7 but less than 9 years
More than 9 years
109
153
130
27.8
39.0
33.2
Position
Owner
Manager
272
120
69.4
30.6
Number of employees
1-50
51-200
40
352
10.2
89.8
Age of business
6-10 years
More than 10 years
180
212
45.9
54.1
International business
Yes
No
207
185
52.8
47.2
Table 4.4 shows the demographic information of all respondents. A total of 398 (39.8 percent) useable responses were obtained out of the 1,000 questionnaires sent. According to Sekaran (2003) the minimum of 10 percent response from the sample justifies the rational to start and perform the analysis. In this study, the business owner or managers were identified as the key informants. They are considered the most appropriate as they are the best positioned personnel to have the broadest knowledge of the overall issues under investigation.
Based on the data from Table 4.4, the majority of the respondents were male with a total of 280 respondents (71.4 percent), while female respondents accounted for 28.6 percent (112 respondents) from the total sample. The business type of manufacturing had the highest number of respondents with 216 (55.1 percent). Other type (service business) with 176 respondents accounted for 44.9 percent.
130 respondents or 33.2 percent indicated that they have had more than 9 years experience in business, whereas the highest level of business experience was more than 7 years but less than 9 years with 153 respondents or 39.0 percent. 109 respondents or 27.8 percent indicated that they have had more than 5 years but less than 7 years experience in business.
272 or 69.4 percent of the respondents were owners, while 120 respondents or 30.6 percent were managers. From these analyses, it can be concluded that the majority of the respondents were from the manufacturing and service businesses, they were owner or manager and have sufficient knowledge to be in this industry.
89.9 percent or 352 respondents were of a medium size business with between 51 and 200 employees, whereas 10.2 percent or 40 respondents were of a small size business with less than 50 employees.
212 respondents or 54.1 percent indicated that their businesses have been established for more than 10 years, 180 respondents or 45.9 percent established from 6 to 10 years.
Finally, 207 respondents or 52.8 percents are international businesses, and 185 respondents or 47.2 percent are domestic businesses.
4.3.5 Univariate Normality
Univariate normality computation is conducted using z-scores of skewness statistics, standard error of skewness as well as kurtosis statistics to analyze the dataset. The z-score of skewness more than 3 needs to be transformed since it is considered as non-normal data (Hair et al., 2010). This study found that the absolute of minimum and a maximum skewness value indicated normal distribution because the value z-score is below 3 (refer to Appendix G). Therefore, data distribution is individually normal.
4.3.6 Multicollinearity Tests
The multicollinearity was examined using the variance inflation factor (VIF) measures and tolerance value. The result of VIF and Tolerance for all variables under study as shown in Table 4.5 found that VIF values were ranged between 1.586 to 2.239 that were below the threshold value of ten (<10) and tolerance value of all variables were ranged between 0.447 to 0.631 that is substantially more than 0.10 (Hair, Black, Babin, & Anderson, 2009). The VIF and tolerance of the variables indicated that there was not evidence of severe collinearity between the independent variables.
Table 4.5
Multicollinearity Test – Variance Inflation Factor (VIF) and Tolerance
Variables
Collinearity Statistics
Tolerance
Variance Inflation Factor (VIF)
Continuous Learning
.561
1.784
Inquiry and Dialogue
.536
1.865
Team Learning
.575
1.738
Embedded System
.598
1.671
Empowerment
.524
1.909
System Connection
.608
1.644
Strategic Leadership
.494
2.024
4.3.7 Reliability Test
Cronbach’s alpha reliability analysis is applied to test the internal consistency to respect multi-dimension. Nunnally (1967) and Shih and Fang (2004) minimum Cronbach’s alpha of reliability is sufficient above 0.6 for the early stage of the research.
Table 4.6
Descriptive Statistic of Reliability
Variables
Number of items
Cronbach’s alpha
Composite reliability
Continuous Learning
3
0.871
0.975
Inquiry and Dialogue
3
0.889
0.981
Team Learning
3
0.917
0.983
Embedded System
3
0.923
0.984
Empowerment
3
0.865
0.978
System Connection
3
0.923
0.984
Strategic Leadership
3
0.929
0.984
Organizational Innovativeness
11
0.934
0.991
Organizational Performance
9
0.920
0.989
As shown in Table 4.6 and Appendix H, each construct is Cronbach’s alpha reading with particular attribute to related ranging from 0.865 to 0.934.
Composite reliability is listed in Table 4.6. There is acceptable value that indicates a possible internal consistency. The calculation of composite reliability is based on the standardized factor loading resulted from the final modified structural model (refer to Appendix H). The formula of composite reliability is shown as follows:
Composite reliability = (∑ standardized loading) 2
(∑ standardized loading) 2 + €∑j
With respect of composite reliability and Cronbach’s Alpha value, George and Mallery (2003) indicated that reliability greater than 0.9 is considered to be excellent, greater than 0.8 is good, greater than 0.7 is acceptable, greater than 0.6 is questionable, and less than 0.5 is poor. Additionally, Sekaran (2000) indicated that the closer the reliability gets to one the better it is. Therefore, with a value between 0.983 to 0.991. The composite reliability of constructs are of an excellent standard. It is considered that any scores above 0.60 is of an acceptable standard (Nunnally, 1967). All reliability in this study are above the acceptable standard.
4.3.8 Validity Test
Validity refers to the accuracy of measurement, whether the conceptual and operational definitions are truly a reflection of the underlying concept to be measured (Burns & Bush, 1995; Neuman, 1994). There are three types of validity: content validity, construct validity, and criterion validity.
Content Validity
Content validity was used to represent the degree of accuracy between a set of measures and the concepts of interest (Cronbach & Meehl, 1955; Hair et al., 2010). Prior to launching the survey, the questionnaire was pretest and pilot test to validate content validity and was generated first in an English version and then translate into Thai. The process for the pilot test as described in Chapter 3 (refer to page 108) included two researchers and two owners from a sample in the pilot test. Their role was to ensure the clarity of each question. As a result, five questions were modified and included in the final questionnaire.
Construct Validity
Construct validity was used to confirm that the indicators aligned with the factors as they are measuring instrument adequacy (Schwab, 1980; O’Leary-Kelly & Vokurka, 1998; Cronbach & Meehl, 1955). This study performed construct validity for nine measurement constructs, that represent the theoretical latent, they are continuous learning, inquiry and dialogue, team learning, embedded system, empowerment, system connection, strategic leadership, organizational innovativeness, and organizational performance. Construct validity was evaluated by convergent validity, discriminant validity, and nomological validity.
(1b) Convergent Validity
Convergent validity is established by high factor loadings and high significant levels of the indicator variable (Schwab, 1980). In order to assess convergent validity, it is essential to evaluate the statistical significance of the estimated parameters between constructs and their items. This study used Confirmatory Factor Analysis (CFA) (refer to Table 4.7) to estimate the values of the factor loading between indicators and the factor. The factor loadings indicate the correlation between the indicators and the factor. The values of the factor loading should be greater than 0.50 for each indicator in the factor (Hair et al., 2010). A single-factor CFA is carried out only when feasible, given that CFA needs at least four items per latent variable to obtain degrees of freedom. When this condition was not achieved, the corresponding construct was allowed to correlate to another construct to obtain factor loadings (Anderson & Gerbing, 1988). Consequently, a single factor model was performed for organizational innovativeness and organizational performance, whereas the learning organization factors (continuous learning, inquiry and dialogue, team learning, embedded system, empowerment, system connection, and strategic leadership) were correlated.
Table 4.7
Convergent Validity – Confirmatory Factor Analysis (CFA)
Exogenous
Variables
Code
Indicators
Factor
Loading
Continuous
Learning (CL)
CL1
In my organization, people help each other learn.
0.818
(3 items)
CL2
In my organization, people take time to support learning.
0.849
CL3
In my organization, people are rewarded for learning.
0.828
Inquiry and
Dialogue (ID)
ID1
In my organization, people give open and honest feedback to each other.
0.910
(3 items)
ID2
In my organization, whenever people state their view, they also ask what others think.
0.800
ID3
In my organization, people spend time building trust with each other.
0.859
Team
Learning (TL)
TL1
In my organization, people have the freedom to adapt their goals as needed.
0.876
(3 items)
TL2
In my organization, people revise thinking as a result of organization discussions or information collected.
0.939
TL3
In my organization, we are confident that the organization will act on our recommendations.
0.845
Embedded
System (ES)
(3 items)
ES1
My organization creates systems to measure gap between current and expected performance.
0.899
ES2
My organization makes its lessons learned available to all employees.
0.897
ES3
My organization measures the results of the time and resources spent on training and learning.
0.886
Empowerment (EM)
(3 items)
EM1
My organization recognizes people for taking initiative.
0.854
EM2
My organization gives people control over the resources they need to accomplish their work.
0.862
EM3
My organization supports members who take calculated risks.
0.769
Table 4.7 (Continued)
Convergent Validity – Confirmatory Factor Analysis (CFA)
Exogenous
Variables
Code
Indicators
Factor
Loading
System
Connection (SC)
SC1
My organization encourages people to think from a global perspective.
0.888
(3 items)
SC2
My organization works together with the outside community or other outside resources to meet mutual needs.
0.918
SC3
My organization encourages people to get answers from multiple locations and perspectives when solving problems.
0.877
Strategic
Leadership (SL)
SL1
In my organization, leaders mentor and coach those they lead.
0.869
(3 items)
SL2
In my organization, leaders continually look for opportunities to learn.
0.930
SL3
In my organization, leaders ensure that the organization’s actions are
0.912
Organizational
Innovativeness
OI1
Management actively seeks innovative ideas.
0.809
(11 items)
OI2
Innovation is readily accepted in program/project management.
0.751
OI3
Technical innovation, based on research results, is readily accepted.
0.762
OI4
Innovation in this business unit is perceived as too risky and is resisted.
0.761
OI5
In new product and service introductions, our company is often first-to-market.
0.734
OI6
Our new products and services are often perceived as very novel by customers
0.745
OI7
In comparisons with our competitors, our company has introduced more innovative products and services during the past five years.
0.771
OI8
In comparisons with our competitors, our company has lower success rate in new products and services launch.
0.773
Table 4.7 (Continued)
Convergent Validity – Confirmatory Factor Analysis (CFA)
Exogenous
Variables
Code
Indicators
Factor
Loading
Organizational
Innovativeness
OI9
We are constantly improving our business processes.
0.719
OI10
Our company changes production methods at a great speed in comparison with our competitors.
0.711
OI11
During the past five years, our company has developed many new management approaches.
0.707
Organizational
Performance (OP)
OP1
In my organization, return on investment is greater than last year
0.780
(9 items)
OP2
In my organization, sales growth is greater than last year
0.769
OP3
In my organization, average productivity per employee is greater than last year.
0.648
OP4
In my organization, time to market for products and services is less than last year.
0.773
OP5
In my organization, take care for customer complaints/needs is better than last year.
0.764
OP6
In my organization, the cost per business transaction is less than last year.
0.680
OP7
In my organization, market share is greater than last year.
0.811
OP8
In my organization, the profit volume is greater than last year.
0.773
OP9
In my organization, the additional capital is greater than last year.
0.752
Order Now