374 APPENDIX B. INTEGRATION ON ROUGH PATHS∗

Theorem B.0.3 Let ai,bi ∈ R. Then for p≥ 1,(n

∑i=1|ai +bi|p

)1/p

(n

∑i=1|ai|p

)1/p

+

(n

∑i=1|bi|p

)1/p

Proof: First note that from the definition, p−1 = p/p′.

n

∑i=1|ai +bi|p ≤

n

∑i=1|ai +bi|p−1 (|ai|+ |bi|)

≤n

∑i=1|ai +bi|p/p′ |ai|+

n

∑i=1|ai +bi|p/p′ |bi|

Now from Lemma B.0.2,

(n

∑i=1

(|ai +bi|p/p′

)p′)1/p′( n

∑i=1|ai|p

)1/p

+

(n

∑i=1

(|ai +bi|p/p′

)p′)1/p′( n

∑i=1|bi|p

)1/p

=

(n

∑i=1|ai +bi|p

)1/p′( n

∑i=1|ai|p

)1/p

+

(n

∑i=1|bi|p

)1/p

In case ∑ni=1 |ai +bi|p = 0 there is nothing to show in the inequality. It is obviously true. If

this is nonzero, then divide both sides of the above inequality by (∑ni=1 |ai +bi|p)1/p′ to get(

n

∑i=1|ai +bi|p

)1− 1p′

=

(n

∑i=1|ai +bi|p

)1/p

(n

∑i=1|ai|p

)1/p

+

(n

∑i=1|bi|p

)1/p

B.1 Finite p VariationInstead of integrating with respect to a finite variation integrator function F , the functionwill be of finite p variation. This is more general than finite variation. Here it is assumedp > 0 rather than p > 1.

Definition B.1.1 Define for a function F : [0,T ]→ R

1. α Holder continuous if sup0≤s<t≤T|F(t)−F(s)||t−s|α <C < ∞. Then from this inequality, it

follows that |F (t)−F (s)| ≤C |t− s|α .

2. The function F has finite p variation if for some p > 0,

∥F∥p,[0,T ] ≡ supP

(m

∑i=1|F (ti+1)−F (ti)|p

)1/p

< ∞

where P denotes a partition of [0,T ] ,P = {t0, t1, · · · , tn} for

0 = t0 < t1 < · · ·< tn = T

also called a dissection in this subject. It was also called a division when discussingthe generalized Riemann integral. |P| denotes the largest length in any of the subintervals. It will be always assumed that actually p≥ 1.

374 APPENDIX B. INTEGRATION ON ROUGH PATHS*Theorem B.0.3 Let a;,b; € R. Then for p > 1,h 1/p n 1/p n 1/p(z bro) < (x wi") + (z mt)i=l i=l i=lProof: First note that from the definition, p—1= p/p’.n n 1VY lai tbil? < ¥ lai +bi\?~ (jail + |bil)i=l i=ln i n Oi< y a; + h;|?/? la;| + y |aj +bjl?/? [bi|i=l i=lNow from Lemma B.0.2,n Np I/p' 7 1/p h sp Wp 7, I/p< (5 (mir) (Zia) (% (lai+bil?/”’) (E15)i=]h 1/p! h \/p h \/p= (Zie-+0i") (z i") + (E15)i=l i=l i=lIn case Y°"_, |a; +b;|? = 0 there is nothing to show in the inequality. It is obviously true. Ifthis is nonzero, then divide both sides of the above inequality by (Y'_, |a; + bi|”) 1/P' to geth ly Ch \/p h \/p h \/p(x roi) = (z i) < (x i") + (x mi") |i=l i=l i=1 i=lB.1 Finite p VariationInstead of integrating with respect to a finite variation integrator function F’, the functionwill be of finite p variation. This is more general than finite variation. Here it is assumedp > Orather than p > 1.Definition B.1.1 Define for a function F : [(0,T] ~R1. & Holder continuous if supy<s<;<r FOfollows that |F (t) — F (s)| <C|t—s|®.<C <.. Then from this inequality, it2. The function F has finite p variation if for some p > 0,mn \/piRlyorn=op(Srn-Feor) <=i=1where Y denotes a partition of |0,T|, FY = {to,t,-++ ,tn} forO=t0 <t <-+<th=Talso called a dissection in this subject. It was also called a division when discussingthe generalized Riemann integral. |Y| denotes the largest length in any of the subintervals. It will be always assumed that actually p > 1.