21st May 2020

---

Functional cFDR results (chromosome 22)

---

From my MCA analysis:

• Dimension 1 explains 51% of the total variance
• Dimension 2 explains 35.5% of the total variance
• Dimension 3 explains 4.6% of the total variance
• Dimension 4 explains 4% of the total variance
• Dimension 5 explains 1.5% of the total variance

This implies that we should be iterating over several dimensions.

---

Iteration 1: Functional cFDR on original p values and -q1 (to ensure low q enriched for low p) to obtain v1.

Iteration 2: Functional cFDR in a two-step approach to deal with big positive and big negative q being enriched for small p. (1) (v1, -q2)=v2_tmp (2) (v2_tmp, q2)=v2. Note that cor(v2, q2)=0.00802974 here so this method wouldn’t automatically be selected.

Iteration 3: Functional cFDR on v2 and -q3 (to ensure low q enriched for low low p) to obtain v3.

Iteration 4: Functional cFDR on v3 and q4 to obtain v4.

---

Below, I plot V1,…,V4 against the original p values: