Understanding consumer online shopping behaviour from the perspective of transaction costs

 

---

206
5.2 DATA CLEANING AND SCREENING 
During the data collection process, a total of 984 online shoppers completed the survey. The 
raw data obtained from the questionnaires underwent preliminary preparation before they 
were analysed using statistical techniques advocated by Kumar et al. (2002). Data cleaning 
and screening were performed to check for incompleteness and inconsistencies in order to 
ensure accuracy and precision of the data.
Data cleaning includes consistency inspection and treatment of missing responses. Missing 
responses represent values of a variable that are unknown, either because respondents 
provided ambiguous answers or their answers were not properly recorded. An examination of 
basic descriptive statistics and frequency distributions were conducted to screen the data set.
First of all, the case with the same response number selected (e.g., all the questions were 
answered with number “5”) were removed from the sample since it meant that respondent did 
not answer the questions seriously. As a result, 4 cases with the problem of inconsistency 
were deleted from the original sample (984).
In addition, as suggested by Sekaran (2003), the questionnaire should be discarded if 25% of 
the items in the survey have been left unanswered. Thus, of the 980 competed surveys, 18 
cases were deleted due to the omission of more than 25% of the responses, resulting in a total 
of 962 responses being used for the data analysis. Nevertheless, of the remaining 962 cases, 
21 cases contained a small number of missing responses randomly distributed throughout the 
surveys. According to the recommendation by Hairs et al. (1998), these cases were examined 
to determine if patterns existed in the missing data. As there were no apparent patterns in the 
missing data, an imputation method commonly used within survey research (Kamakura and 

---

207
Wedel 2000, Myrtveit and Stensrud 2001) whereby the missing value is estimated based on 
values of other variables, was the remedy chosen to handling missing data, i.e., missing 
values were replaced via linear interpolation using SPSS.