respondents with 31 of questions and the outcome is valid, so the researcher could continue the research.
b. The research held in BNI, kantor besar, Sudirman, Jakarta. c. The respondent chooser by this research, following the order; active
customer, means come to BNI minimum once a month, have fix income, the respondent is chooser with simple random sampling
methods.
D. Data Collection Methods
These research methods did on October 2008 used collection and observation data, data is collected from BNI Jakarta, there are:
1. Primary Data: data is getting from the survey result with giving
questioner to the respondent that suitable with the population characteristics of customer BNI Sudirman, kantor besar, Jakarta.
2. Secondary Data: the general data is getting from the last research and
literature about relationship between customer loyalty and customer satisfaction.
Data that already collected by the respondent, and then will be selected, edit, suitable with the researcher necessary.
E. Analysis Methods
This research method using the analysis methods is the descriptive methods. The purpose is to get some figure of descriptive or systematic outline,
actual, accurate about the fact, also the phenomenon relationship that will be researched and explore.
The level of essential of customer value factor and customer satisfaction towards customer loyalty will do by Likert method. The instrument of the
question will end result the total score for every member of sample that represent by every score that already write down, on the below:
Table 3. 1 Likert Scale
Likert Scale Score
Strongly disagree Disagree
Neutral Agree
Strongly agree 1
2 3
4 5
Source: Freddy Rangkuti 2003
1. Validity Test
Validity is an indicator that shows the level of validity or legitimacy of an instrument. An instrument is valid if could measurement what do
we need. Valid means that instrument could be use to measure what we want to calculate. Validity that used in this research is construction
validity is the framework from one concept; with calculate among correlation with moment product correlation formula. With r = 0, 01 if
less from that so the statement is not valid.
2. Reliability Test The instrument that used for many times to measure the same object
will make a same data of the result. To measure the reliability coefficient could use with alpha cronbach formula.
Where: R
11
= Instrument Reliability k = sum of Question
σ
2
= Total Variance σ
2
b = Sum of Variance 3. Multiple Linier Regression
− =
2 1
2
1 11
σ σ
b
k k
r
In this research using model is multiple linier regression analysis. This model used because wants to looking for the influence value X1 and
satisfaction X2 to customer loyalty Y in BNI. Multiple linier regressions the formula is:
Y = α + b
1
X
1
+ b
2
X
2
+ ε Where:
Y = Dependent Variable a = constantan price Y if X = 0
b = Coefficient Regression X
1
= Independent Variable Customer Value X
2
= Independent Variable Customer Satisfaction ε = Standard Error
4. Classic Assuming Test a. Autocorrelation
Autocorrelation is the relationship that happen among the member from observation that systematize in time combination that connected
with time length that could means that, correlation relationship from every variable now will have same condition with future condition.
One of analysis that used to analyze the autocorrelation is with Durbin Watson
− −
=
n n
n
e e
e d
2
1
Where: Ho: there is no correlation
Ha: there is positivenegative correlation if Durbin Watson -2 so there is positive autocorrelation, and if DW 2 so it is negative
correlation.
b. Multikolinearity Multikolinearity used to prove there is correlation linier between
independent variable in regression model. If independent variable creates the perfect correlation it could be perfect multikolinearity.
To analyze is there any multikolinearity or not in regression model with do the observation the tolerance value and variance inflation
factor VIF. Independent regression model multikolinearity have VIF value approximately on 1 until 10 and have tolerance number close to
the fault in estimation independent variable capacity as a dependent variable predictor.
The multikolinearity test do by consider VIF value regression if VIF value of independent variable bigger than 10 so, the conclusion
that a variable have strong linier relation with another independent variable. To decrease the multikolinearity is disappear the independent
that variable from the similarity regression.
c. Heterokedastisity
Heterokedasity show that the variance of variable there is no same for every observation. The fault of heterokedasity happen not random,
but show the systematically relation based on the biggest one or more. Heterokedasity have purpose to check is there any regression
model happen in differentiation variance of residual or another observation is still same so can called by homokedasity and not
heterokedasity. To analyze is there any heterokedasity, use some method, which are:
i Observe the graphic plot between variable prediction variable connected ZPRED with the residual SRESID. To detect is there
any heterokedasity could do with consider current method in scatter plot graphic between SRESID and ZPRED, where point Y
is Y that already predicted and point X is the residual Y prediction – Y real.
ii Basic analyzes, if there is current point that make a systematize figure. If there is no clear model by point spread on the above and
below on zero in Y point, so there is no heterokedasity.
5. Hypothesis test a. Partial test
The objective to make a conclusion regarding the influence of independent variable X to dependent variable.
So, the formula such as:
Sb Se
b t
test
− =
Χ −
Υ =
n Se
Sb
2 2
2
2
− ΧΥ
− Υ
− Υ
= n
b Se
α
Where: a = Constantan
b = Coefficient correlation N = Total sample
Sb = Error Coefficient Correlation Se = Estimation Error
If T test T table so Ho rejected and Ha accepted, means independent variable by partial have significant influence to dependent
variable . If T test T table so Ho accepted and Ha rejected, means
independent variable by partial don’t have significant influence to variable dependent.
b. F test simultan
To do hypothesis analysis, so there is role that needs to be concerned is the formulation zero hypotheses ho and we also need to put the
alternative hypothesis ha, such as: 1. Ho : ρ = 0 there is no significant influence between variable X1
and X2 to variable Y. 2. Ha : ρ ≠ 0 there is significant influence between variable X1
and X2 to variable Y. This analyzes to analyze all the independent variable together can
influence to dependent variable. So the formulate to F test is:
1 1
1 2
− −
− =
K n
R k
R f
If F test F table so Ho rejected and Ha accepted, means independent variable together have significant influence to dependent variable.
If F test F table so Ho accepted and Ha rejected, means independent variable together don’t have significant influence to dependent variable.
F. Research operational variable