Validity and Reliability Test Classic Assumption Test

53 a. Other relatively stable characteristics of the individual that influence the test score, such as intelligence, social desirability, and education. b. Short-term or transient personal factors, such as health, emotions, fatigue. c. Situational factors, such as the presence of other people, noise, and distractions. d. Sampling of items included in the scale: addition, deletion, or changes in the scale items. e. Lack of clarity of the scale, including the instructions or the items themselves. f. Mechanical factors, such as poor printing, overcrowding of items in the questionnare, and poor design. g. Administration of the scale, such as differences among interviewers. h. Analysis factors, such as differences in scoring and statistical analysis.

1. Validity and Reliability Test

The research instrument questionnaire that both must meet the requirements of the valid and reliable. To determine the validity and reliability of the questionnaire should testing performed on the questionnaire by using validity and test reliability. Because the validity and reliability aims to test whether questionnaires were distributed to obtain research data is valid and reliable, then for it, the author will also conduct the second test against research instruments questionnaires: 54 a. Reliability According to Malhotra 2004:267, reliability refers to the extent to which a scale produces consistent result if repeated measurements are made on the characteristic. The value of variable reliability demonstrated by the Cronbach Alpha coefficient. A variable is said to be the Alpha Cronbach coefficient of Reliability when 0.60, when the variable is said to be 0.60 not reliability. b. Validity According to Malhotra 2004:269, the validity of a scale may be defined as the extent to which differences in observed scale scores reflect true differences among abjects on the characteristic being measured. According to Ghozali 2006:49 the validity test is used to measure the validity of a qustionnaire.To significant test is done by comparing r count with r table for degree of freedom df=n-2. In this case n is sample amount. df=20-2=18 and alpha 0.05 we get r table 0.444 two tail test. To test the validity of each questions, it can be seen from Corrected-Item Total Correlation coloum. If the score of Corrected-Item Total Correlation r table the questions is valid.

2. Classic Assumption Test

Before conducting hypothesis testing, in accordance with the provisions of that in the multiple linear regression test should perform the classic assumption test prior to testing errors on the regression model used 55 in the study. Therefore, at least four basic assumptions must be met that would normally be tested by multiple linear regression Santoso, 2012: 221 a. Multicollinearity test According to Ghozali 2006:95, multicollinearity test aims to test whether the regression model found a correlation among the independent variables customer service X1, store design and display X2, communication mix X3, location X4, merchandise assortment X5, and pricing X6. Good regression model should not happen correlation between the independent variables retail mix Consist of customer service, store design and display, communication mix, location, merchandise assortment, and pricing, If among the independent variables correlated with each other, then these variables are not orthogonal. To detect the presence or absence of multicollinearity among the independent variables in regression model, it can be seen from Tolerance and VIF value. Cutoff value commonly used to indicate wheter there is multicolliniearity or not is Tolerance value ≤0.10 or equal to VIF value ≥10. b. Normality Test Normality test aims to test whether the regression model, the dependent variable customer satisfaction and independent variables variables customer service X1, store design and 56 display X2, communication mix X3, location X4, merchandise assortment X5, and pricing X6 both have a normal distribution or not. If the distribution of the residual values can not be considered to be normally distributed, then it is said there are problems with the normality assumption. According Ghozali 2006: 149, the principle of normality can be detected by looking at the spread of the data dots on the diagonal axis of the graph probability plots or by looking at the histogram of the residual. Basis for decision making as follows: 1 Detection of the histogram, if the normal curve in the graph follow a bell shape, then the data are normally distributed. 2 While the detection of the normal probability plot on the graph, if the data dots spread around the diagonal line, and follow the direction of the diagonal line, then the regression model to meet the assumption of normality. If the spread of the data points do not follow the direction of the diagonal, then the regression model did not meet the assumption of normality. This statistical test that can be used to test the normality of the residuals is a statistical test of non - parametric Kolmogorov- Smirnov KS Ghozali, 2006: 151. Basis for decision making, when the value of the Kolmogorov Smirnov significance greater than 0.05, it can be said to be normally distributed data. If the value 57 of the significance of the KS test is smaller than 0.05, it can be said the data was not normally distributed. c. Heteroscedasticity test aims to test whether the regression model of the residual variance occurs inequality an observation to other observations. If the variance of the residuals one observations to other observations stable, it is called different homoskedastisitas and if it is different called heteroscedasticity. Good regression models is that happened homoskedastisitas or did not happen heteroscedasticity Santoso: 2012: 238. 1 Looking at the scatterplot graph, if forming certain patterns, such as dots form a certain pattern regularly wavy, widened then narrowed, then heteroscedasticity indicates has occurred. If there is no clear pattern, and the points spread above and below the 0 on the Y axis, then there is no heteroscedasticity Santoso, 2012: 240. 2 Glejser Test Glejser test is done with the regressed absolute value of residuals against the independent variables customer service X1, store design and display X2, communication mix X3, location X4, merchandise assortment X5, and pricing X6. Guidelines from glejser test is looking at the significance level of each independent variables customer service X1, store design and display X2, communication mix X3, location 58 X4, merchandise assortment X5, and pricing X6 on the dependent variable customer satisfaction. If the significance level yield number 0.05, it can be said regression model does not contain any heteroscedasticity. Ghozali, 2006: 129.

3. Multiple Linear Regression Analysis