topic Re: Two Sample Paired T Test Results in Statistical Procedures

Two Sample Paired T Test Results

avalenzuela6 — Mon, 08 Oct 2018 20:43:08 GMT

I am trying to compare two different versions of the same product. Since they are independent of each other, I decided that I should do an independent sample t-test. When I go to do a two-sample test on SAS OnDemand for Academics, the results don't make any sense. When I do a paired test however, it generates perfect results. Have I decided to run the wrong test or am I just missing something?

Re: Two Sample Paired T Test Results

PaigeMiller — Mon, 08 Oct 2018 21:54:20 GMT

I have looked at your PDF file, and this requires more explanation, specifically which test is which, and which ones make perfect senses and so on.

Can you also show us the code you used?

Furthermore, since you state that your data is not paired, you should not do a paired t-test, it will be meaningless.

Re: Two Sample Paired T Test Results

Rick_SAS — Mon, 08 Oct 2018 22:57:57 GMT

Show the code, please. This looks like the output from PROC UNIVARIATE and it looks like a BY group analysis in which each group has one or two observations.

Re: Two Sample Paired T Test Results

avalenzuela6 — Thu, 11 Oct 2018 00:03:49 GMT

The file is what happens when I run a two sample t-test. Code is below. I guess what is confusing me is that it doesn't show any graphs when I do two sample as opposed to the paired test, which does.

* Task code generated by SAS Studio 3.7

* Generated on '10/10/18, 5:00 PM'

* Generated by 'avalenzuela6@capellauniversity.edu'

* Generated on server 'ODAWS01.ODA.SAS.COM'

* Generated on SAS platform 'Linux LIN X64 3.10.0-693.21.1.el7.x86_64'

* Generated on SAS version '9.04.01M5P09132017'

* Generated on browser 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_13_6) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/69.0.3497.100 Safari/537.36'

* Generated on web client 'https://odamid.oda.sas.com/SASStudio/main?locale=en_US&zone=GMT-07%253A00'

ods noproctitle;

ods graphics / imagemap=on;

/*** Test for normality ***/

proc univariate data=WORK.IMPORT normal mu0=0;

ods select TestsForNormality;

class SF;

var UniDun;

run;

/*** t Test ***/

proc ttest data=WORK.IMPORT sides=2 h0=0 plots(showh0);

class SF;

var UniDun;

run;

Re: Two Sample Paired T Test Results

Rick_SAS — Thu, 11 Oct 2018 11:50:51 GMT

I suspect that the SF variable contains more than two levels. The SF variable should contain two values that indicate the product version. The UniDun variable contains the measurements. Your data should look like this example:

SF UniDun

1 1.23

1 3.21

1 4.23

1 2.14

...

2 4.53

2 3.43

2 3.61

2 3.87

From the PROC UNIVARIATE output, it looks like your data might look more like this:

SF UniDun

1 1.23

1 3.21

2 4.23

2 2.14

...

12 4.53

12 3.43

13 3.61

13 3.87

You can use PROC FREQ to produce the values of the SF variable:

proc freq data=WORK.IMPORT;
tables SF;
run;