Identically distributed random variables #
Two random variables defined on two (possibly different) probability spaces but taking value in the same space are identically distributed if their distributions (i.e., the image probability measures on the target space) coincide. We define this concept and establish its basic properties in this file.
Main definitions and results #
IdentDistrib f g μ ν
registers that the image ofμ
underf
coincides with the image ofν
underg
(and thatf
andg
are almost everywhere measurable, as otherwise the image measures don't make sense). The measures can be kept implicit as inIdentDistrib f g
if the spaces are registered as measure spaces.IdentDistrib.comp
: being identically distributed is stable under composition with measurable maps.
There are two main kinds of lemmas, under the assumption that f
and g
are identically
distributed: lemmas saying that two quantities computed for f
and g
are the same, and lemmas
saying that if f
has some property then g
also has it. The first kind is registered as
IdentDistrib.foo_fst
, the second one as IdentDistrib.foo_snd
(in the latter case, to deduce
a property of f
from one of g
, use h.symm.foo_snd
where h : IdentDistrib f g μ ν
). For
instance:
IdentDistrib.measure_mem_eq
: iff
andg
are identically distributed, then the probabilities that they belong to a given measurable set are the same.IdentDistrib.integral_eq
: iff
andg
are identically distributed, then their integrals are the same.IdentDistrib.variance_eq
: iff
andg
are identically distributed, then their variances are the same.IdentDistrib.aestronglyMeasurable_snd
: iff
andg
are identically distributed andf
is almost everywhere strongly measurable, then so isg
.IdentDistrib.memℒp_snd
: iff
andg
are identically distributed andf
belongs toℒp
, then so doesg
.
We also register several dot notation shortcuts for convenience.
For instance, if h : IdentDistrib f g μ ν
, then h.sq
states that f^2
and g^2
are
identically distributed, and h.norm
states that ‖f‖
and ‖g‖
are identically distributed, and
so on.
Two functions defined on two (possibly different) measure spaces are identically distributed if their image measures coincide. This only makes sense when the functions are ae measurable (as otherwise the image measures are not defined), so we require this as well in the definition.
- aemeasurable_fst : AEMeasurable f μ
- aemeasurable_snd : AEMeasurable g ν
- map_eq : MeasureTheory.Measure.map f μ = MeasureTheory.Measure.map g ν
Instances For
In a second countable topology, the first function in an identically distributed pair is a.e.
strongly measurable. So is the second function, but use h.symm.aestronglyMeasurable_fst
as
h.aestronglyMeasurable_snd
has a different meaning.
If f
and g
are identically distributed and f
is a.e. strongly measurable, so is g
.
This lemma is superseded by Memℒp.uniformIntegrable_of_identDistrib
which only requires
AEStronglyMeasurable
.
A sequence of identically distributed Lᵖ functions is p-uniformly integrable.
If X
and Y
are independent and (X, Y)
and (X', Y')
are identically distributed,
then X'
and Y'
are independent.