turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- SAS Survey Predictive Analysis

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

06-07-2016 01:02 PM

Dear Team ,

I have recently started working on a new survey data which has data points for the last 1 year. The customers can rate us on a scale ranging from 1-10 where 1 being the lowest and 10 being excellent.

I have performed the EDA/ descriptive part of the contributing variables as to which are most prominent, region level, day of the month etc and other variables and have come up with a list of most significant variables.

Now i am not sure how to perform a prediction based on these as it is the first time i am working on a survey data. One option i can think of is creating clusters from these variables however then the same has already been achieved using the freq tables.

Can we run a logistic regression on these 8-9 predictor variables i have and the outcome var i.e. C-sat can be categorized in 3 levels similarly we have for the NPS scores say multinomial logistic regression.

Any thought & inputs would be of great help.

Regards, Shivi

Accepted Solutions

Solution

06-13-2016
09:36 AM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Shivi82

06-10-2016 01:45 PM

A non-parametric measure of association should be considered first:

```
Proc corr data=myData Pearson;
var rating1-rating4;
with ratingFinal;
run;
```

PG

All Replies

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Shivi82

06-07-2016 01:10 PM

What is EDA?

PG

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to PGStats

06-07-2016 01:23 PM

Exploratory Data Analysis (EDA), a term popularized by Tukey.

https://en.wikipedia.org/wiki/Exploratory_data_analysis

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Rick_SAS

06-09-2016 12:24 PM

HI Team,

Any suggestions. Would a chisquare test do justice.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Shivi82

06-09-2016 05:01 PM

What are you trying to "predict"? Are you trying to estimate one of the 1-10 rankings given other charcteristcs? Something else you have a measure of, say total $ sales, based on the rankings? One ranking element using other ranking varialbes?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to ballardw

06-10-2016 01:01 PM

So we get C-sat results from across the globe. There are 4 parameters on which we get rated: for instance- a) knowledge of the representative, b) understanding and control as a customer service rep and then other 2 parameters as well. Then we have 1 final rating on which we get marked called as Final rating and it is not a sum of the other 4 as i highlighted above.

So i need to test whether there is an association between the Final rating and the other 4 parameters.

Say for example, there would be a scenario where in we get rated 9 or 10 on all of these 4 parameters where 10 is highly satisfied however on the Final rating it could be a 1 or 2 which is extremely dissatisfied due to uncontrollable issues such as any bug in the system or a new feature result in the CRM app.

So my question was whether we can run a chi square test of association and see if there is any significant relationship.

Please suggest.

Thanks

Solution

06-13-2016
09:36 AM

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to Shivi82

06-10-2016 01:45 PM

A non-parametric measure of association should be considered first:

```
Proc corr data=myData Pearson;
var rating1-rating4;
with ratingFinal;
run;
```

PG

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Email to a Friend
- Report Inappropriate Content

Posted in reply to PGStats

06-13-2016 09:39 AM

Thanks for the solution. I think this is the perfect start to see how these var are related to the final var. From here on i can build on with the most significant var if any.