Research on Retail service quality and customer loyalty is being carried out presently by me.
Brief Description of the Study:
In our Scope, Retail Service Quality is considered as the independent variable. Further, it has five dimensions or Variables, Which is measured by 22 questions or items (Physical Aspects, Reliability, Personal Interaction, Problem Solving and Policy).Meanwhile, Customer Loyalty is considered as the Dependent Variable, which is measured by 8 questions or items.
Research Methodology
First of all, I intend to perform the EFA to factorize the items of independent variable (22 Questions), which will give the unique and uncorrelated factors under the independent variable.
Meanwhile, I don’t want to perform the EFA for whole 30 question or items. In which, 8 items are used to measure the dependent variable. Generally, there is a coloration or association between independent and dependent variables. Therefore, I only used 22 items under the independent variables to perform the EFA to construct the unique model.
Secondly, I wish to perform the CFA with the identified items of the factors under the independent variable via EFA and items of dependent variable, those are not factorized via EFA ( 8 items) to confirm the whole set of model , which contains both independent and dependent variable under the measurement model.
Thereafter, I also like to perform the Structural Equation Model (SEM) to find out the path way between independent variables (Physical Aspects, Reliability, Personal Interaction, Problem Solving and Policy) and dependent variable as Customer Loyalty.
Sampling Adequacy
What is the required level of Sample size for the advance statistical tools as Exploratory Factor Analysis and Structural Equation Model?
Whether Two thousand (2000) sample size is enough or too much or less to get the significant model
If it is too much, how we will tackle it smoothly to get the Significant Model.
Hi, Good Morning
You want to do research means, first you should download some journal articles related to your research or study. Based on that reference you have to find the adequate sample size.
Suppose your study related to prevalence, then you can use the prevalence (in %) and find the sample size with the use of following formula:
4pq
n = --------------
d2
otherwise, you should go through that references and based on that you have to find the sample for your related study. This is the correct procedure.
Kindly do it. All the best.
Thanks.
Thank your sir; anyway, simply, i know each and every step which should be followed in the research.
My Questions are,
Can we put the all the items (Items of independent and dependent variables altogether to factorize the model in one base via EFA?
If Not, Please Justify It
Can we put the all the items (Items of independent and dependent variables altogether to factorize the model in one base via CFA?
If Yes, Please Justify It.
Can we perform EFA before we confirm the model via CFA?
Whether sample size as 2000, which was derived with mathematical formula from the population is enough or less or more for the advance statistical analysis in terms of EFA and Structural Equation Model ?
In response to sample size: The first thing FA will do is to discard all surveys that are incomplete. You can recover some of these by imputing the data. Search internet using key word "imputation," and also check out this site: http://www.missingdata.org.uk/.
If you are still at 2000 surveys, the next question is what was your population size? Is this about every person in Sri Lanka, or about the 2500 employees at a local business? The model is only as good as the sampling protocol. If you want to describe cell phone use by city dwellers in Sri Lanka, then you need a sufficient sample from enough of the cities. 2000 surveys from Colombo may not tell you much about the people in Jaffna or Galle.
Another consideration is what is the goal and what are your data like? If you want predictive accuracy of an average value and your data are well clustered about the mean, then life is good. If your data is skewed, with heavy tails, then you might be better off considering a maximum accuracy method. This is especially true if you want overall predictive power for both the mean and outliers. See ODAjournal.com. I liked this article (http://optimalprediction.com/files/pdf/V2A3.pdf). You may also find http://optimalprediction.com/files/pdf/V2A38.pdf relevant. The ODAJournal web site has a book "Optimal Data Anlaysis" that directly links to Amazon.com where the book can be purchased. You should contact Paul Yarnold (he has a research gate account) directly to make sure this program is right for you. This approach has many good applications, but it is also not for everyone.
Did you do a correlation test using all of your initial 22 independent variable items, and what was the cut-off value for deciding whether two questions are related or not, as well as what questions to include in the finale EFA. This is an important first step to avoid issues of collinearity. I imagine you should also know that what type of correlation test to be conducted depends on whether the question responses represent discrete or categorical variables. Secondly, you cannot include both independent variables and dependent variable questions in one EFA. The whole essence of your EFA is to identify the factor dimension and therefore constructs measured by the pool of independent variables items. Thereafter, you should use these factors in the model against the dependent variables. So why would you want to use both independent variable questions and dependent variable questions as components of your EFA.
Secondly, you mentioned that you have 8 questions representing your dependent variables. Does that mean you have 8 independent variables or how do you want to categorize the dimensions of your dependent variables? One good place for you to start is to review the scientific literature with regards to your research subject and see what dimensions have been used to address the questions you are asking. This creates better justification and rationale for your decisions. EFA is only a tool to start, but the questions you load should first be driven by the science. This helps guide your decision on what questions should go together in formulation of the dependent variables and the factor dimensions of independent variables. That way, your research output becomes reproducible.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size