#### RachelSSF

##### New Member
I have data collected with 5 Likert Scale Questionnaire to establish factors influencing customer satisfaction using online banking. The factors included Credibility, Efficiency, Ease of Use, Security, Problem Handling and Product/Service Portfolio. A total of 6 independent variables. In this case, what is the best hypothesis test recommended? Can I still perform Multi Regression using Customer Satisfaction as Dependent Variable? Because my questionnaire does not measure customer satisfaction. Therefore, I don't have any data available for customer satisfaction. What is the best analysis to perform and interpret this data? After performing the normality test, all my 6 variables are not normally distributed.

#### noetsi

##### No cake for spunky
I am confused what you mean by the fact that you do not measure customer satisfaction. If you do not measure it you can't analyze it. What is your dependent variable.

There is an ongoing dispute if you can or can not use likert data as interval data in the dependent variable, justifying linear regression (which is what I think you mean by multiple regression, formally this means all regression with more than one predictor).

Ordinal logistic regression is better according to many in the social science realm, but in economics linear probability models are run for two level dependent variables in part because its a lot harder to interpret ordinal logistic regression (or even binary logistic regression). Slopes are not intuitive in logistic regression, economist accept the errors that such methods generate to get easier answers (and better diagnostics IMHO).

What are the five levels of the dependent variable.

#### Karabiner

##### TS Contributor
A total of 6 independent variables. In this case, what is the best hypothesis test recommended? Can I still perform Multi Regression using Customer Satisfaction as Dependent Variable?
As noetsi observed, you cannot analyse a dependent variable if you do not have data on that dependent variable.
What exactely was the research question for this study?

With kind regards

Karabiner

#### RachelSSF

##### New Member
SPSS Data

I have a study to establish the factors that influence customer satisfaction with different internet banking services. Customer Satisfaction and Attributes to satisfaction measures using 5 Likert Scale total of 23 Questions. Credibility (5 Questions), Efficiency (5 Questions), Ease of Use (3 Questions), Security (3 Questions), Problem handling ( 5 Questions) and Product/Service Portfolio (2 Questions). So, I have 6 independent variables here. I am confused about how to compute my dependent variable which is Customer Satisfaction.

#### Karabiner

##### TS Contributor
You do not need to write essentially the same again, this time with bold typed words.
Did you chose the research question on your own, or was it given to you? If the latter,
who gave it to you?

Elsewhere you gave a link to the study questionnaire. There it says:
This study is expected to give an overview on key dimensions to be considered by
banks to increase the adoption and quality of IBS. In order to do this, there are few things need to be
considered such as
1- Demography of potential customers (family income, race, education level, etc).
2- The preference internet banking provider.
3- Attributes of customer satisfaction.
4- Factors influencing the internet banking usage

So, AFAICS credibility, efficiency (etc.) are the dependent variables here, they are different aspects
or facets of consumer satisfaction. You can analyse them separately, or you can somehow combine
them to achieve a measure of "total satisfaction".

And you want to know whether factors such as family income, race, educational level etc., and also
the primary internet banking service used, are factors which are associated with a client's consumer
satisfaction. Can we say so?

How were the data collected, and who large is your sample size?

With kind regards

Karabiner

Last edited:

#### RachelSSF

##### New Member
You do not need to write essentially the same again, this time with bold typed words.
Did you chose the research question on your own, or was it given to you? If the latter,
who gave it to you?

Elsewhere you gave a link to the study questionnaire. There it says:
This study is expected to give an overview on key dimensions to be considered by
banks to increase the adoption and quality of IBS. In order to do this, there are few things need to be
considered such as
1- Demography of potential customers (family income, race, education level, etc).
2- The preference internet banking provider.
3- Attributes of customer satisfaction.
4- Factors influencing the internet banking usage

So, AFAICS credibility, efficiency (etc.) are the dependent variables here, they are different aspects
or facets of consumer satisfaction. You can analyse them separately, or you can somehow combine
them to achieve a measure of "total satisfaction".

And you want to know whether factors such as family income, race, educational level etc., and also
the primary internet banking service used, are factors which are associated with a client's consumer
satisfaction. Can we say so?

How were the data collected, and who large is your sample size?

With kind regards

Karabiner
Hi karabiner, the questionnaire was given by the university. We need to collect data using the questionnaire provided with a sample size of 100. In the data analysis part, we are expected to perform descriptive statistics, normality tests, parametric/nonparametric and correlation/regression analysis.

#### Karabiner

##### TS Contributor
If this is a beginner's task, then you can just perform descriptive statistics, bivariate statistics
(for example: does the mean satisfaction score differ between men and women [t-test]) and
correlations (for example: Spearman correlation between income category and satisfaction),
and linear regression (for example: prediction of satisfaction by service used, gender,
family income, and education level).

What you have to decide: whether you want to perform each of the bivariate/correlational/
regression analyses six times (on each of the 6 satisfaction scales), or only once on the total
satisfaction score. If you want to do the latter, you'll have to determine a way to aggregate
the 6 single scales to 1 total scale. Did they give you a hint how to do that? Often, such
scales have already been used before, and sometimes there's manual containing such
information.

With kind regards

Karabiner

Last edited:

#### RachelSSF

##### New Member
If this is a beginner's task, then you can just perform descriptive statistics, bivariate statistics
(for example: does the mean satisfaction score differ between men and women [t-test]) and
correlations (for example: Spearman correlation between income category and satisfaction),
and linear regression (for example: prediction of satisfaction by service used, gender,
family income, and education level).

What you have to decide: whether you want to perform each of the bivariate/correlational/
regression analyses six times (on each of the 6 satisfaction scales), or only once on the total
satisfaction score. If you want to do the latter, you'll have to determine a way to aggregate
the 6 single scales to 1 total scale. Did they give you a hint how to do that? Often, such
scales have already been used before, and sometimes there's manual containing such
information.

With kind regards

Karabiner
Hi Karabiner,

The confusion sort out. Actually, there are some mistakes made in the questionnaire. Missing questionnaire for customer satisfaction. As a student, I don't really wish to challenge the authenticity of the questionnaire until I found the facts from expert people like you all. Finally, it was admitted that the questionnaire itself had the problem. I managed to complete the task. Thanks for your reply and help.