How to use simpleM result?

[kayeee_]

Hello. I am trying to analyze over 300 SNP(single nucleotide polymorphisms)s data

I came across information suggesting the use of corrected P-values (MultipleM) for SNP data's P-values.

I found a code implementation in the provided URL (9 pages - 3.2 Alternative to Proc PSMOOTH),

https://support.sas.com/resources/papers/proceedings09/240-2009.pdf

The example result is shown below.

7 parameters explained 99.5% of All parameters.

How should I set the corrected P-value?

Is it 0.05/7, 0.05/340, or is there another approach?

Proportion	Cumulative	Meff	alphaG
0.5193	0.5193	.	.
0.2424	0.7617	.	.
0.1547	0.9164	.	.
0.001	0.9174	.	.
0.005	0.9224	.	.
0.0154	0.9378	.	.
0.0572	0.995	7	0.007143
0.0004	0.9922	8	0.00625
0.00004	0.99224	9	0.005556
…	…	…	…
0	1	300	0.000147

[sbxkoenk · Accepted solution]

Hello,

I looked at page 9

https://support.sas.com/resources/papers/proceedings09/240-2009.pdf

SAS Global Forum 2009 -- Paper 240-2009
Beyond SAS/GENETICS™
Amina Barhdadi, Yassamin Feroz zada, Marie-Pierre Dubé
Montreal Heart Institute and Université de Montréal, Montreal, Canada

The code is clear.

... ...
 If cumulative>0.995 then Meff=number; /* cumulative and number are two provided variables in eigenvalues data set*/
Run;
/*Step4-Apply the Bonferroni correction formula to adjust point-wise significance level*/
Data Meff;
 Set Meff;
 alphaG=alphaE/Meff; /*alphaE is the experiment-wise error rate*/
Run; 

Meff is an estimate of the effective number of independent tests.
In your use case Meff = 7 , so you divide the test-wise error rate by 7 to get the experiment-wise error rate (Bonferroni correction). Like 0.05 / 7 = 0,007143 .

Important Note: 
The last comment line says : /*alphaE is the experiment-wise error rate*/

The experiment-wise error rate is on the left-hand side of the equality sign of course.

What you divide by 7 (Meff) is the test-wise error rate.
So, the comment is wrong (should say alphaG) ... or alphaG and alphaE should switch places.

Koen