Programming the statistical procedures from SAS

Need robust method to calculate statistical difference

Accepted Solution Solved
Reply
Occasional Contributor
Posts: 9
Accepted Solution

Need robust method to calculate statistical difference

Hello, 

I am new to data analysis and need help with the following.

I have a survey data summary table where i have results from state and area of that state. I am interested in yes response

It looks like following for each question 

                   Participants          Yes               C.I                   No           C.I.  

STATE         7450                    88%          83%-93%          12%     10% -14%

Area             668                       82            80%-84%          18%    16% -20%

 

How do I calculate statistical difference in above two and decide if its statistically significant?

I can do it by calculating yes and no response number from percentages given and do chi square test but it wont give a robust results.

Really appreciate input and help.


Accepted Solutions
Solution
‎04-30-2018 03:40 PM
Occasional Contributor
Posts: 9

Re: Need robust method to calculate statistical difference

You don't need to perform any statistical calculations. As long as CIs don't overlap, you can conclude that the results are statistically different. In the example in question, the result is statistically significant as CIs don't overlap.That is how governments interpret data. Correct me if I am wrong

View solution in original post


All Replies
Super User
Posts: 23,773

Re: Need robust method to calculate statistical difference

Chi Square is robust...and your N is large enough to obtain reliable estimates.

 

If you want to explore non-parametric methods, look at PROC NPAR1WAY.

 


GARYV wrote:

Hello, 

I am new to data analysis and need help with the following.

I have a survey data summary table where i have results from state and area of that state. I am interested in yes response

It looks like following for each question 

                   Participants          Yes               C.I                   No           C.I.  

STATE         7450                    88%          83%-93%          12%     10% -14%

Area             668                       82            80%-84%          18%    16% -20%

 

How do I calculate statistical difference in above two and decide if its statistically significant?

I can do it by calculating yes and no response number from percentages given and do chi square test but it wont give a robust results.

Really appreciate input and help.


 

Solution
‎04-30-2018 03:40 PM
Occasional Contributor
Posts: 9

Re: Need robust method to calculate statistical difference

You don't need to perform any statistical calculations. As long as CIs don't overlap, you can conclude that the results are statistically different. In the example in question, the result is statistically significant as CIs don't overlap.That is how governments interpret data. Correct me if I am wrong

Super User
Posts: 23,773

Re: Need robust method to calculate statistical difference

Please do not post the same question multiple times. Note that I've merged these into a single post and moved them to the Statistical Procedures forum as that seems better suited.
SAS Employee
Posts: 386

Re: Need robust method to calculate statistical difference

See the SURVEYFREQ procedure in SAS/STAT software.
Occasional Contributor
Posts: 9

Re: Need robust method to calculate statistical difference

Posted in reply to StatDave_sas

HI , Thank you for reply. I only have output table with percentages  and C.I.. I don't have all the raw data set to use SURVEYFREQ procedure.

Super User
Posts: 23,773

Re: Need robust method to calculate statistical difference

All you need is the counts or you could recreate the data in an approximate fashion if you have the 2x2 numbers.

 


GARYV wrote:

HI , Thank you for reply. I only have output table with percentages  and C.I.. I don't have all the raw data set to use SURVEYFREQ procedure.


 

☑ This topic is solved.

Need further help from the community? Please ask a new question.

Discussion stats
  • 6 replies
  • 199 views
  • 2 likes
  • 3 in conversation