2/3/20

T h m : Matt metical Te s t f o r Independence

events A t B o r indepandet i f
1 ) PCAIB) -PLA) pegg
2) P(BIAS = p (B) esaiudat
3) P # aB ) = PCA, y}

e) An A r B independent?
P (BIA) ='s independent
P(B) Is =
]

f) a r B t c independent?

P(Btc) = #
] not independent PCelB)
#Pee)
P( B ) , Is±
P (Bnc) PCB)P( C )

Thu: t c u o f told probability

Suppon eats E , , Ez, Ez,..... En, fa m a petition
1) F-i n Rg 0 Hi Ff
=

:÷i÷÷÷÷÷.si#*Oi&::.U(
K

EnnAD
me.
= PCE,n d ) t PLBanA) t . . . t HENNA)
=
PLEDPLAI#t PLEDPCA l e f t PCEn) PLAIED

Thin: Bayes thm

Supper E,,....,En form a partition
then for any element A (PLA) 40)
P (Eita) -
PCEints
Puts
[Link]#AEi)-
PLE,)PlAlE,)t....tPlEn)PlAlEn)

ed
/¥§§$
"%, -344¥,
aw
Ba → 212 1 W

133 → 413 I V
R . C, a bowl then R-S. -chip
,

a) f .t o prob set Rdrnip

you
a

PCR) PCCB,nR)u (Bank) u l 133N R))

=

" e "
PCB, N R ) t PCBs'nR)t PlB3nR)
=

=
PCB,)P(RIB,) t PCBs) P(1211322%133)(R#
=
(b)( t ) a E ) (E) tf's)(F)
b) given the chip was R find prob thati t
c a m e from 133
P
(BIR) =
[Link]#=P(Bz)(RlBs)-
(3%4/5)=[Link]#P42)-a -

PlBzlR)=PCpBq,zgR)-
=PlBz)lRlBa)- MR)
←
EYE) = .
40
X
p(13,112) = . 12

Price problems: PCB,), P (Ba), PCBs) -1 / 3

posterior pot : PCB, IR), P (Belk), P (133/12)

ed boy hes 513 t 4W marbles in Lpodet

413 f s U marbles i n 12 pocket
R, S, a marble from L t t r a s h to 12 pocket
R , -5. m-04 from R

"Gio
find prob o f
a) B

m..,÷÷÷¥E:
P(3cnBr)tP (Vin Bp)
= P(BL)P(13,2113,)t P( Y ) P (13,211 =

( F ) (fo) +
(9)(Fo) -Toto
 \

b) P(BolBr) - P(BLnBr)
-
= Plz) PCB, IBC)
P (Bp) P(Be)
2 § ' fo =-610
¥

2/5/20
AT Lawrence Livermore Act, Lab.
1 0 , 0 0 0 employees 5 o r
spies
Use polygraph t o catch spies
prevalence = P(5) =
g¥o =
n o o o s

sensitivity = P (tls) e -91

specificity =
Pf-Is') e . 94

a) find prob of f . I n ( f ) positive

P (Hse) -
l- pl-lse) e -06

b) find prob o f f . l u A negative

PC-Is) l - P (tls)
a = . 09

c ) find prob 1 2 . 5 . employee t e s t t

'¥00.
Ptt) =
Plsnt) t p (int)
=
pls> Plots) tP(so)(P(pls.)
-9995 (iou) 060425
(00051.91) t
= .
 Bayes t h i n

Plstt)
d) [Link]#=DlslPI=- =

Pct)
=.looos&
- 060225 = .
00753
e) P l s ' t ) -[Link]#=Plse)PI=E99kI
f . Pf) 060925
=
0,99995

Random variables

Def: A R - V. i s a function belt assigns a Fiber to

each outcome i n
your
$
experiment)
X- R , U. (changes from expirmard t o

X = number (positive value o f X)

Def: A discrete R i v. c a n assume a finite o r

countably I number of values

Def: a continue, D .U . c a n assume a n y
value i n a n

interval (unecuntably-)

ed
#
§ flip a coin twice
§ =
{Htt, H t , t H , TT,}
let X = # H
a ) what i s "support o f X"?
§ × - {0,1, 2
b) find prob o f mass function X
(prob distribution)

7 : o I 2
fly:(3) t
Spx) = P (X--x): I ¥ ¥
7=91,2

six#ft..
Note: suppose X is a
discrete D .U .

1)
flax)> o V x E$x
2) E f(x) a t
XE$×

e y -
e) Bou' h a l chips
z → "I'
y → "y"
172.5. 2 chips v/o replacement

let X a
sum of th 2 dress
a) find flu of X
7 : 4 L 8
I ✓
k$-PEDE#¥ =

FIX)=P(x.4 ) =
Plz, n 2) ' =
Plz,)P(2,12,)
=
I o o
} = To
3%1=4)
f(b) a

=, PG?,n4z)t PM,n2z))
P (2, n
42) t P l 4 , n z )
=
Plz) P l 4212,) PM,) Plz, 14,) t

E . Is + ¥.}=¥
f l 8) P ( * 8)
= =
p(4,n%) -
Pla,)P(4,14,):
" ÷ :

4¥.
6) P/xzllxE6) - P ( x z b , #
a x e 6)
=Pp{¥¥ =,'g4¥-8/92.889

c) L t X-product o f 2 d a s

x : 4 8 16
f l y ) = P(x.x ) : E ¥ ¥

2/7/20
D e f : cumulative Distribution function Gdf) o f a

discreet 12,U . X is
C-( x ) = D (XE x )

ftp.j-szxcuhcnlimffx)
: 2

- O

* 1 -s
him F ( x ) : I

× A s
if 7 E y , the, C-( x) E
fly) →
calf is

handeersing
exJ suppose X has pms
f/×) = p (x:X) - I t
7 =
1, 2 , 3

a.
÷¥¥.
a)
[Link]"
-1 I:&:3

[io¥ d o

¥.
i x ± s

f) Plez), F (2) I
=

c) P(12 XE 2) F b ) -f-( i )
= =
I - l =}
d) M k 2) i
2/10/20
Def: suppose X is a discrete 12.0. with
1) expected value of X (man) i s
µ = E(x): Exfed
XE$x
* long r u n
average
2) if h (x) is a function o f X , then
F- CtlxD=&hCxKf6s
YES!
3) the verienee of X is 0 2 = Venue(x)
* disoles viability =
E [(x-NI]
units'
A =
{(x-µ)2fCx)
¥ typical squared devotionfun [Link],
4) standard deviation o f X
0 =
SDK) =
I t
* describes variability
¥ units
¥ topical deviation frm the m a n
exJ Manufacturer sells battery 2 -pales
let Xi # low voltage
know t h e pmt is
× : o I 2

fly): Plax): 7 -2 . I

a) µ . - f-(x): Exfld
51=0-
D E

C. 7 )t I (2)t 2,.fi)
= 0.4

f)
÷t¥.
find p i
02 = Ucr (x) =
E ((x-µ)2]
= E (xyz) -FG)
(O--4) ( I - ,4)' t . 2
2
e- . . ? t 7

12-'4)
'
. . I =
0 . 4 4 bettors'

c) find 0
k spa) E d = T . m e 0.663 batteries

d) and E(×')
Eli)=§.gg?ifH)=0?l.Dtl4.Dt2Y.i)
= . 6
 Thin: Compddiard formula far berione (x)
the o a r of a R, V . Y c a n be found a s

Ocr (o) Ecw) =

-p-= f-( x ) (EG)] -

(discrete card:
proof i n
Uab): Eflx-µ))
=
{(X-µ)' f(×)
= E (x' - 2 µ × t µ2)f/×)
e { x ' f k ) -2µ£×fG)tµ2 E f f )
= E l l i ) - 2 µ % t µ' toot
i
f ki) - µ '
= Efi) -

(Eu
e) Ucr (x): E C I ) -
(Eve =
,
b- . 43.49

Det: Discrete Uniform Distribution

The
R. V . X has discrete [Link] destination i a f
eal o f t h values for f'× has equal probability
is! (x,.... %) N c o
f l y ) P( X2 7 ) -ht
= for 41×43... - h e
exJ suppose X distr.
has
l,...., ' 0
a disc unit. o n

f l y ) : P l a x ) to X - 42,3...- 1 0 =

a) P (X> s ) . p (x-6) t . . . . . t h x : O ) S .

b) µ E M : E x f f x )
=

=¥÷x. to
=
to (172 t . . . . -Ho) =
n(nz#
 t .to # E
=
5.S
e ) r - # = E E ) (E(xD" -

Eli) Ex-flo) =
38,[Link] ,

÷÷÷÷÷:
=

II, [Link] -8.25

=
to (1427.... to')

÷.→)
Binomial Distribution

D-f ! Har h [Link] B a r trails where prob o f

success I eat trial is
p. I f
× . # successes i n the rituals
thn X has a binomial [Link]/ parameters n i p
X - B i n (nip)
2/12/10
Thm: X - Bin (h,p) i f f far [Link].... n

f t p . p l x e x ) elf)p×(l-p)""

S, S, f si, f
- - - - - -

T T 3- n - I h

note: Binomial expansion

forconstants a t 6, (at6 )" = ¥ (f)a " b""

(atb)" =
a ' t abt-62
(atb)? If (f) a x 62-×
of(3) a ' f?
gf})a2f2-2
(f) <o f?
- 0 - I
=

= a' + a b t 62

Note: let a
= p , b-t-p

*
&Wp"(tp)"" = (p-h-p))"- I
 Thm.. if X has a 8 i n (n,p) then
1) E ( x ) a
np
2) War( x ) = n
p (l-p)
proof
=
Eld-§, 7 . (Y) p"(I-p)""
=
§, 7 5 1 1 ¥ P "(Ip)'

= IE,
I t y p" (Lp)"

E. ¥ j ? ¥ *, : p " ups""
- "' " '
=
np

"y= p - i ,
""
§, (II) p " f-p)"
- " '

=
rip m = n-I

=
[Link]/j)pYf-p)m#
=
rip
exJ roll die 1 0 times
Let X : # 6 ' s
find P (11=2)
Old
Way:-
P#2) = P (b, b. bi.... i)
t . . . .

t . . . .

Binomial Way:
X-Bi n (ni lo p-te)
P (11=2) -
(9) pill-p)"" ' o - 2

= (Y) (f)24. g)
4 5 . # (E)
=

8=0.29 07

¥
2:(lo-2'll

¥
yrs:
ed Tires
9% a r e detective
(indep)
5ps. 3 0 tires 12,5. a r e

let X i #defective
X- B i n o m i a l (n.-30 pi-02)
a ) PLY.-3) ( f )pill-p)" e

(330).023
(l.. oz)
30-3
c -

= 4060 (oz)' (. 98)"

:O. 0188

f) p (x? 1)
=
, Phil)tP(x. 2) Paid
l - P( x u )
. . . .

= l - p (x:o)
= . 4545

c) On average
how
many
(3×
defective
tins do w e expat?
F-(x) =
If X . 0) -oz" ([Link]""
=
np
c - 301.02)
i . 0.6
d) Var (x) enpll-p) -0.588 def

e) SD(x) T = H =
0.7L i d e f.
2/14/20
Binomial
Def: Negative Distribution
Have a
sequence o f independent Barn. trick
ul prob o f s a u c e s p . suppose,
X . t r i a l o n which get r t h success
than X-NB (rip) =
( I I ) f l i p ) " fo r
x=r,rH,
Thm: X - N B (rip) i f f . . . . . . . . . .

flies = P ( X i x ) ,
S, f

¥ ¥ ¥ ÷ .". ÷ ÷
Thai ¥. ( I I ) [Link]" - I

Thmi. I f X - NB (rip), then

1) E (x)
- F
2) Va. (x), r ¥
p-
 Prob h i t target
e) a
o n
any
shot =
0.4 2
Assume slots o n
indep.
let R .V. X = shot in which hit target fo r
the 3rd time
X - NB ( r=3, A . 92)
a) find prob h i t tarsal for 3rd the a f t shot

P(* s) =
( I I ) pr ( t p ) ' - 3
=
( { 1 ) o,yz3([Link])'
=
( E ) C.42131,58)-= i tso

f) o r
average hit 3 r d time a n relief shot?
E ( x ) : =p =
¥ = 7, 1 9 shots
2/11/20
Negative binomial

Note: distin
a geometric is a NB [Link]/r=l
Y r Geo (p)

Note; a binomial dist ' n is analogous to sampling

with replacement
*
independence does not change
a
p

Def: type, geometric dist ' n -analogous to sampling

without replacement
Acu N diets
*
M Buceesses
* N-M n o r success
D.s, objects
ulo replament
n

I f X-# success t h X - (n,N,M)

--Plax).-1%(1-11)
Pot: X- H b In,N,M) iff ff)
7 = 0 ,1 , . . . . , n (Nn)
X E M

a - M E N-M
Thai. I f X- H G (n,N,M) then
1) E (x) n # = like "rip" ~

2) Verb): n # (I-E) (II)

-
likenp(i-p)

suppose bout 20 ATM's 4 blue

12.5. 3 do repleement

let X =
# B
a ) find prob of 0 Blue
old fashion
uey:
P(x.o ) . P(Bin Bin 135)
= P(Bi) P(Bil B,')P(135/13,413)

¥ =
¥ . if 0.491o r =

AG
way:
X-Hb (n =3) (N-a)(May)
Pl*ol=µ§÷÷)_M¥3
1.991
6) P(x=2). vey
old
= P [(B, n B a r 133')u (B, B i n 133)u (BinBank)] n

= P (B, n 132 nB5) t . . . t ( )

P

=µµI¥µ}¥=
=

x-Hol#2)
rosy

c) Expected number h Blu

Elx) =
n# = . 3 . I = . 6

Note: Wen N i s "large" d i n i t "shell", w e can

approximat
e
Hb (n,N,M) by a B i n (n, p . F)

G) 120 applicant, 80 a r e qualified

12.5. 5 applicants To repleement
l e t RU. X H qualified =

find P(11=2)
Y - H 6 (n:S, N i k o , 14=80)

[Link]#.=h5fI:jI--o"
X : Binomial (n=s, p=¥ a } )
(1-3)%0.165
Plez) -

(f)p×(t-p)"- × . (f)#
Hyper geometric last thing a n tat

study!
 Def: Poisson Distin
spse X i s a R. V, such that X - # o f o c c u r s

i n a n interval o r
region where
e) = expected # o f occurrences i n the intruder

region
assume 1) # of occur rences i n honour-lapping intrels o r
resins
a r e
indup.
2) . . .

3) . . .

Def: X - Dois (s) iff

f-(x) = Play) = e ¥ for *
0,1,...

Note:.÷e÷÷-e".÷÷¥ ' I i
'

1) E l x ) - I
2) Vaulx), a
§ a
geiger counter i s hit 5 times per s e e ,
o n
average
Pois assumptions hold.
t x # o f hits
= i n 1 see

X - Poi, ( 5 = 5 )
a) find prob of 0 hits i n l sea

[Link]?#e-so,1 e . o o h

6) h i t s £ times i n 1 see

Pba8) e-s 58e

= . 06 5
8 7
c) p(Xz2) =
P(xF2)tPlx )....
e l - Dlx-725
= I - [[Link]) t PHD]
= . 9596

d) Elx) =
des
e) Proo o f 0 hits i n lanin
D= # of hits i n 1 min
X-Pisf ) solo),
= 300
5.148×10""
Plxeo) =

e}¥o° =
2/21/20
Review
Form 30% orange 98% frost damaged
I
~
o r

70% apple 0 11 i l i ;

knw Ploroge) a . 3 pfflo) = - 4

P (apple) = . 7 P(HA) I = .

a) Plitt) l - p l a t ) . -9
=

b) P(Fn A) = P(A) DAIA) = . > x . I 0 7 = .

c) p(f) partition p(onf ) tP(Anf)

±

= D( o ) P(Flo) t PLA)P(f I A)
-
3 x i t t . ) x . I = 19

d) Platt)
.

=
Bars them
=D,¥¥ = MAIL.t#..7
x I .

" 8

pathos-ma,?
- '

e) intio):
=
Dlf)tp(o) -
(Plo) P l a n e
- 19 . 3 -
( . 3 4) x . =
. 3 7

f) D .[Link] 5 trees [Link]

2o r

X= # f
[Link]/
n=Spil..l9)Pfx=2)
(f)[Link]"
1915-2
(f),
1 92/1.. = #
 (indep)
g) repeatedly select trees at random
find prob that loth tree i s 1st frostdanged
tree A t
X = frost
X m N B (r--I,p=.la) k Geo([Link])
Dlx.-lo) (ki)p' (tp)"
- r
-

-fie:).la'll..")"'
-
819x . 1 9 =
- ?

h) repeatedly select trees at random

fird prob th. lott tree i s 4th frost danged
Ye tree 4 2 ' ferl
D= NB (v. 4, p...19)
Plx-1011981).la" ([Link]"" = = ?

i) Ar. f t A indep? MAH) # (A)

Bob (FIA) = , I not indepedy Planf ) + PLASMA
P lF ) = . , g )

j) super 10 trees, 4 ol l i d o r apple

12.9. S trees u/o replant

K. # A
X - A G ( n i 5,10=10, M-4)
M¥3)
-l:µ÷¥=µ}¥-e?
ID spse 15,000 trees 3,000 Apple
12,5. S o o trees w l o replant
find prob get 9 0 Apple
X v HG ( h s o o N 15,000, 14=3,000
a - =

Photo) (3%0914%93%0)
-

¥ ?
V o x
=

= ,
02428
use binomd approxinle

Ya # A
D r Bin (ni soo,p=
3,88% e .
2)
P (4=90) -
({ooo).,' ( 1.-2) soo-90 = . 02443
exJ bout 312 2W chips
D .S , 2 chips d o replaent
chips sane color → win $1
chips diffont i , → lose $ 1 .
let ke winnings

a) A d of X
pmf
× ; - I I

[Link]#xlP(x=-l)=P/D,nwz)tP(w,tnRD-
ly
=P/l4)Mvzll2,)tPlw,)P/R,/U,)
} . f t } . § = 0.6
P ( * /) = P (Rink) + P l u , nth) z -4

b) [Link]?
fpix:Yn=(%feiIi.4xzl
c) EH)= § thx). -if, b) t 114) =
- - 2

d) Elex), sexff): (b) e' (4) = - 7 .

'
e- t

e) Volk): EA-ND-Stx-µYf6)
=
f (x).IE/x)Y
- f . 2)
2
= / = -96
f) to (1×1)=0
exs 23 students
→ 49144 F)
a) number o f dies lost 4 t h 4 f e t the beginning

¥ b 19£
M

b) Randomly permute the list

f i r d prob all 4 female, o r in the fief 4 spots

4:19'.
2 3 , 7
Ye H F in l't 4 slots
X - H b (n=4, N-23, 14=4)
P(x-4)
 3/2/20
Def: Exponential Distri
Spa w e hear a Poisson Process v/ m e n

Nk 470 if
between
X= waiting t u r n occarners

then Xn Exp ( s )
when the X
p d f of i s

81×5:{de-* t o o

O.W
o

¥° I

1) [fad, =
The-" d x - - e -' l ?
°
= O t e° =/

2 ) F-( x ) . I
3) v o l t s = ¥
4) e d f = f l a t Plxen) =

{,?,-¥0,,,
 Calls o w n at a call center according

to a poi's prooeess w / m e n of e) = }
calls per min

a) fund prob of 2 calls i n next min

Ye # of all, i - I w i n

X - Pais ( X-I)
DLX.2) e¥(bI
=e}¥ I i = = . 0398

b) find prob waiting t u n e fo r n e x t call i s

less than to win
X: waiting t i m e fo r next * K i n n i h

V a
exp( r ) : b) →
Jk): di# x s o

='s e -' 5 " x > o

" X <lo) - "f I e - I x ,

-Es-e-÷ two
= -
e t I
= . 9143
Using this
X, H of all, i n 10 w i n
X m Pais ( i . '%)
PIX?1 ) l - P ( x a l l
i

= l-Plt:o)

¥
l-e-193 / '§
÷

÷ . 914)
c) mean waitin, time between
calls?
Eld ¥=
I , = 3 imin
=

d) find 9 5
t h
percentile o f the waiting t i m e
find × s u c h t h t
-95 = P(pets
=
go" be-It I f
=- e - I t to
=-e-t "t I
e d → e- ×
a , o g

→ -Ix-blocs
→ x = - 3 d, f. os) = f .9 9 m L

than.. Lack o f memory

properly
if X n exp( d) t h
D l x- t , t [Link]/xst.)
e) PIX> 701 X > bo): Mx-10)

Def: Normal Distr-

X- N l µ , o') i f f
f (x) =
1 - e
'Il:*)' - • e x c y

* • I
-

2µL A

"" "°""
¥iE¥E"
Thm. D I, ¥ 0 e-±(¥5'd
x =
/

2) E L A m e

3) Oar( o ) = O-

Y) O =
SD
I a s X - N µ-100, 02=162

E l i
only accepts
[Link].p-
people wl I Q in top 2 %

[Link]
@[Link]
X i loot 20544=132.8
ed weights o f bags of apples
X - N (µ, o - e . a ) lbs
i f 2 . 5 % o f bags weigh less than 4 . 7 lbs
find µ ,

a
A l t
l
4.) N

0.02s i p (x c
4,7)
=p
4 ¥ . 4¥)
=
plz c
"¥0
a-1.96
-1.46 f .2)- 4 . 7 = - N
=
5.092 l b s
Thm .'
Spee X is a random Variable & C i s a
constant

1) F-(ex)
(Eb) =

2) Vortex): C' Volx)

proof: F-(ex) =

[Link]/
If
xldo=cElx)
Vortex).-[Link]-Ela.))
=E[(ex-CENT]
- E l i (x-EM
'

=iE((x-Elx)))
: c ' V e (x)
 Propagation of e r r o r

Reall i Taylor series

acid is indefinett, [Link]
spse
at "a" then
alxle Wait 0¥ final"
-a)
t¥¥a" t...
-

> i

(an Utd by 6 - a )
approx
4 ) [Link]
Note: let fustian of R.V.'s by
approx
a
similar argument
than; spa X -
' l , ,o : )
Taylor
First order approx
about the m e n pix is

Ilka Mpx) + ¥4:*'"'"

nut Elahi)=u(pix)
di. [Link]/aFlx=[Link]'

\
 if µ known i t by
is not replace our
observed odue x

o:-. Volume. #1*+5 ri

(feet)
ed Measure side of cube
L - • (µ, --lo, Oi-i)
Engineering n
1011ft

Volume of cube U = (3

App... Nu ' t
Ou
Elul, lo'
pi tooo
= =

µ .

%:(File.-into?
H i h=m5oi
- - - - - - - - - -
e) Ohms Law / Power

D= I know U - 12 volts
R
Mae resietor i r circut knw
R n . (µ,;D're 5 ) →
every
s ohms

Meason resistance
r e 100

tip--af =
I =,q÷ =
1.94 watts

Of:(k¥1".-miff o'

[Link]#lr=rY
oi=tIlr..Io:
=
¥ o1 '2

Op: ¥ or = 1¥. S - O.o n votts

Approx
power i s
1.4 9 t 0,072 utts
exJ Ohm's Law R= ¥
know
I n -
(a.
of..;))
U - a Cpu, O j a i )
indep.

Measure voltage current

9

ve 100 volts
i - s amps
Approx t
µ, On

Ma#FE ¥ =
÷
=
= 2 0 ohms

[Link]#lu:g:foitlEFlu:am5oI

-HH:*.to:1#l::.;:Io
'

E=hYoitffIoi.
Open i t o u t ¥01
=
}. 2 t '¥-0,2 =
1,2

Resistance is 701 1.2 ohm,

Homework

R t µ , --100, 0,2=42

A: i t 122

knit -
9mi. O i
44001?
y
Assume
V - (per Eu, I)
R-Car, ok.. i )
v.. 1 4 . 4u r : 8.2 ohms

5,17, 22,12