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
- /
- Why no solution computed in PROC POWER

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

09-01-2014 11:00 PM

Dear,

Currently I am using SAS 9.4 to calculate the power with the given size and un-equal variance between 2 groups. However, when I submitted the following code, it turned out there was NO SOLUTION COMPUTED, hence I seek your help on this.

*By the way, I have read the guidance regard to this test method, and found there was no wrong in the code.*

The Code:

proc power;

twosamplemeans test=diff_satt

groupmeans=(1.2 0.1)/*Specify group means: grp1_mean=1.2, grp2_mean=0.1*/

alpha=0.1 /*Specify the test significance level*/

sides=2 /*Specify the test is two-sided*/

groupstddevs=(1.2, 0.1)/*Specify the standard deviation for each group: grp1_std=1.2 grp2_std=0.1*/

groupns=(10 10) /*Specify the number of sample size in each group:grp1_N=grp2_N=10*/

power=. /*Request the power with above prerequisite*/

;

run;

Thanks in advance,

Frank

Accepted Solutions

Solution

09-02-2014
10:28 AM

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

Posted in reply to Jack2012

09-02-2014 10:28 AM

You can use simulation to estimate the power. See the program at Using simulation to estimate the power of a statistical test - The DO Loop but use

RejectH0 = (Probt <= 0.1);

and

method="Satterthwaite".

The estimate of the power is 0.8530, with a 95% CI of [0.8311, 0.8744].

All Replies

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

Posted in reply to Jack2012

09-02-2014 09:42 AM

I was able to duplicate this behavior, and was a bit surprised. I tried two one-sided tests at alpha=0.2 and was able to get results, but am baffled by this particular behavior. It remains until the Group 2 standard deviation gets to about 10% of the Group 1 value. So you could say the power was greater than 0.846.

It appears that this result occurs whenever the ratio of the SD's is greater than 10.

Steve Denham

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

Posted in reply to SteveDenham

09-02-2014 11:59 PM

Dear Steve,

Thanks so much. Yes, this is another good way to present.

Frank

Solution

09-02-2014
10:28 AM

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

Posted in reply to Jack2012

09-02-2014 10:28 AM

You can use simulation to estimate the power. See the program at Using simulation to estimate the power of a statistical test - The DO Loop but use

RejectH0 = (Probt <= 0.1);

and

method="Satterthwaite".

The estimate of the power is 0.8530, with a 95% CI of [0.8311, 0.8744].

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

Posted in reply to Rick_SAS

09-03-2014 12:00 AM

Dear Rick,

Thank you so much, yeah. Simulation is one way to calcualte the power. However, it seems in PROC POWER, this is one bug? Hope in future this can be resolved.

Best,

Frank