diff options
Diffstat (limited to 'issues/effect-size-for-gemma-is-a-must.gmi')
| -rw-r--r-- | issues/effect-size-for-gemma-is-a-must.gmi | 10 |
1 files changed, 10 insertions, 0 deletions
diff --git a/issues/effect-size-for-gemma-is-a-must.gmi b/issues/effect-size-for-gemma-is-a-must.gmi new file mode 100644 index 0000000..137e96e --- /dev/null +++ b/issues/effect-size-for-gemma-is-a-must.gmi @@ -0,0 +1,10 @@ +# Effect size for GEMMA is a MUST + +To: @Zach, @Pjotr +From: robwwilliams +Regarding: Getting a readout of effect size and polarity of effect from GEMMA +Request Type: New feature for GEMMA, User requires these data + +Main Message: Let's Just Do It. Can be done as a kludge. Just plot the beta of the regression equation of the SNP status (ususally two or three states codes as 0, 1, 2 or -1,0, +1) against the phenotype. +The slope is roughly the effect size. This is good enough for most users. The only got-you is slight inflation of due to kinship structure and the Beavis effect when the sample size is low. +QUESTION—can I upload an image from Amelie Baud's talk on effect size? |