Descriptive Statistics: Raw Data

Several things can be done to the raw data in order to see what they can say about the hypotheses (Neuman, 2003). An inspection of the raw data can be done by using the descriptive statistics to find obvious coding errors. The minimum and maximum values for each variable must fall within the admissible range. Pairwise correlations depict that all relationships must be in the expected direction. Meanwhile, listwise deletion of missing values indicates that the data can be used for analysis.

An outlier is an observation that is unusually small or large. Outliers assist researchers in detecting coding errors. According to Bagozzi and Baumgartner (1994), outliers are not recommended to be routinely excluded from further analysis. Data collected were analyzed by using three approaches:

1. Cronbach’s alpha (a) was used to test the reliability. Cronbach’s alpha indicates how well the items in a set are positively correlated to one another. This is to make sure that the scales are free of random or unstable errors and produce consistent results over time (Cooper & Schindler, 1998);

2. Descriptive statistics where the researcher used mean, standard deviation and variance to get an idea on how the respondents reacted to the items in the questionnaire. The major concern of descriptive statistics is to present information in a convenient, usable and understandable form (Runyon & Audry, 1980).

Descriptive summary, including frequency and descriptive, was used to screen the data set. Among basic statistics to use were mean, median, mode, sum, variance, range, minimum, maximum, skewness and kurtosis.

3. Inferential statistics concerned with generalizing from a sample to make estimates and inferences about a wider population (Neuman, 2003...

... middle of paper ...

....e. more than 30 (Hair et al., 2006). Sekaran (2003) suggests the approximation to normality of the observed variables could be investigated by inspecting the data through histograms, stem-and leaf displays, probit plots and by computing univariate and multivariate measures of skewness and kurtosis. Histograms, stem-and-leaf and probit plots indicate the symmetric distribution of variables or sets of variables.

Tabachnick and Fidell (1996) suggest the value of skewness and kurtosis is equal to zero if the distribution of a variable is normal. Chou and Bentler (1995) emphases the absolute values of univariate skewness indices greater than 3 can be described as extremely skewed. Meanwhile, a threshold value of kurtosis greater than 10 can be considered problematic and value greater than 20 can be considered as having serious problems (Hoyle, 1995; Kline, 1998).
Get Access