Help on C Sat Data Analysis - Should I use discriminant analysis?

V

vicky_csam

Have data(Pre collected) from 100 customers for the process services and they are gauged on 10 parameters.Each parameter is being measured on a scale of 1-7,7 being the best.So for each parameter we have 100 feedbacks ranging from 1-7.The Chi-square P value for 10 parameters is 0.5567.

Should I use discriminant analysis?

Pls share your thoughts on analysis for understanding the variance in parameters.Which parameter need to be identified out of 10 and what analysis could be done.

many thanks

V'Dsousa:thanx:
 

Miner

Forum Moderator
Leader
Admin
The correct choice of an analytical tool always relates back to the question that you need to answer (i.e., the hypothesis that you want to test). Can you clearly articulate your hypothesis for us?

You mentioned the Chi-square test, which was not statistically significant, but which variant of the Chi-square and which null hypothesis? If I assume that your null hypothesis was that there is no difference in proportions between groups, then I would conclude that there was no difference in proportions between groups.

Discriminant analysis tests for relationships between explanatory factors (e.g., your parameters) and the customer satisfaction scores of multiple groups. If there is no difference in the scores of the different groups, there is no purpose served by running a discriminant analysis.

Start by clearly articulating your null hypothesis then select a test that is appropriate for that null hypothesis. Chi-square may or may not be that test.
 
V

vicky_csam

Thanks Miner for your reply...

The correct choice of an analytical tool always relates back to the question that you need to answer (i.e., the hypothesis that you want to test). Can you clearly articulate your hypothesis for us?-(--I wanted to find if there is any difference in the parameter.Therefore my hypothesis was that all the parameters are same and the Chi Square value also supported the same.I am not left with any option to disect any specific parameter, had the Chi Square value been significant and individual contribution could be looked to find the significant parameter)

You mentioned the Chi-square test, which was not statistically significant, but which variant of the Chi-square and which null hypothesis? If I assume that your null hypothesis was that there is no difference in proportions between groups, then I would conclude that there was no difference in proportions between groups.

Discriminant analysis tests for relationships between explanatory factors (e.g., your parameters) and the customer satisfaction scores of multiple groups. If there is no difference in the scores of the different groups, there is no purpose served by running a discriminant analysis.

Start by clearly articulating your null hypothesis then select a test that is appropriate for that null hypothesis. Chi-square may or may not be that test.


Help me in disecting my parameters.I have 10 Parameters and I want to analyse these.I cannot do drilling down on all.What should i do?

Thanks in advance.

regards
 

Bev D

Heretical Statistician
Leader
Super Moderator
I might suggest posting an example and articulating what question you have about that data. Unless we clearly understand what question you are asking of the data we cannot offer more specific advice...
 

Miner

Forum Moderator
Leader
Admin
Agreed. Without more information, we cannot recommend any more. I would typically start with graphical data exploration and see where that leads.
 
V

vicky_csam

Have attached the data with little modification the rows are the 114 customers labelled as C1 to C50 & D1 to D64.

The parameters are the specifications for measuring the services and they are rated from 1-7.In some cases the customer has not provided any value to some parameter or Zero in case of NA.the highest rating is 7.

Suggest me your valuable analysis.

regards:thanx:
 

Attachments

  • Data sheet1.xls
    62 KB · Views: 119

Miner

Forum Moderator
Leader
Admin
I was able to perform a preliminary analysis as follows:
  1. A correlation matrix of the 10 parameters. All 10 parameters were strongly correlated with each other. Conclusion: If a customer rates you high (or low) on one parameter, they rate you high (or low) on all parameters.
  2. A Factor analysis. All 10 parameters may be treated as a single factor.
Conclusion from the preceding analyses: Analyze the mean (or sum) of the 10 parameters by customer. Do you have demographic data on these customers, and would you expect any differentiation by demographics?
 
V

vicky_csam

Thanks for your input Miner..

The Customers are from one geography and so not much to stratify on that aspect.

My Intent is to identify the important parameters and make them as Small project Y and do drill down and collect data for factors feeding in that Y.

Will establish relationship between the identified parameters and root cause defined X's.

So still stuck with that.

Analysis at this juncture shows equally important response for all the parameters.The Customers usually work with a stereo mindset and if their experience is good they do not differentiate on various parameters.

Your ideas invited...

Many thanks
 
Top Bottom