UNIT 2

LECTURE 6 : Introduction to hypothesis testing, Type 1 & 2 errors

Hi there,

This is your pal Scooby

Following are my notes for the lecture. These are my handwritten detailed notes. I felt that Prof. R
taught a lot more in the videos than he wrote in the slides. I tried to make them without any
mistakes and cover most of the video material.

But as Beyonce has rightly said : 'To err is human'

I however DO NOT TAKE any responsibilty if you fail your exams following my notes. So please
check the actual lecture slides and videos if you suspect something is wrong.

Also tell me about it at : [email protected]

The notes DO NOT CONTAIN solutions to any homework problems.Only the problems and
exercises solved in the videos are included in these notes.

Here's a motivational quote

'You feel it burn inside you. In your veins, muscles and bones. It rushes throught your body,
penerating your soul.... And finally you are free'
~ Excerpt from Diarrhea diaries

Here's a cute photo I googled

I hope these notes will help you too

Let's get started

Lec # e ; \ntroducdorn to Hypotheax Tes{tn , & Tybel & Typs 2 Etvors 6.tr
p )typothesiE testtnq
h)e noru rno\re. "+vtto-the regiru + scienhftc ezn"lutvX- .Inz wo{ nre do
r,t)Gtl
this ts by Osktn2 binart! luestons obor:k oulc partolneVe,w d. Uike tf
e ts gve,atev *hc.i, tr"lo(z) oF nob, q rr,teqrr(rt) ts :Lero ot nolc. G
ttlustrate th-is r.n.re wtll study a, problenn,.
Whqt ts the -fusfest \da^J to boo.rd a p\ane J
T.ts is ouv queshon *hat hle want to seievr,tt$co\t6 e.vtltdre . h)e o.re
boo'cotrg t'lteraested ttr, tut o vnethod,^g
1. R2F z Re,crv {" t-6
Tne method 6 sirrrple, flrst -$ll *|ne se-ats/Totds tr,n"tlne veav(end.l {lnen
kee,p orr $tting dlt the Jst rol^l r.s$lled.
Z. W1LMA, : \dindo6s, Mr.ddte , Aisle
Trus v,rae*hod .rs sirnpte . €tvst -titL { atU ttne wtr,ndor,rr *z&s, {'in-er, {he
rvrjddte onss cntd thew(,tvtcrttg ttne cdste Aeafs .

5o to coyrctuCt Dry ytseaycl,u rnr,r,n*tl cnttect datA.

7, RzF frv RzF boordi,rng rvt"ethod rnre collected" t*re boording ||me..b
,.

fon 72 jzlbtue +^W. h)e do not Lr-arne o {*t of, o1I the
boardi4g tr,mu but hle h'aw
1z
{.eroge W*e ', 24-2 wuvw ; sLandovd Dev0eHot'. 2.1 naivt*
2.WjLMA,: For wttMA boavdd4g vvt,eth"d hle co(lecbed the boavdingttmr,a
t . 56 Ltuited JUge+. , We do *ol Vr.aw a lr^t o& alt *he
56 boavctt4g tivrrter but rnle h,arne
q yniw ; Stondoyd Savdott% :
{v'evage trvrae : 15. 1.6 tnir,r.s

At?*. 9t * harna G catkd s{ieiernt staHs}icl . \ur,le wt[t rypt stud6

\"'haf that axacttg wLe-or,r\a lauk wg ave saWW 4\,a*, )h,e d"af,q \^'e
ha.ne (ta [nd sto.ytdovd elev. ) ,ts "erfl4tu",rr+ ]o ir,tvestrgate
r^AecvL
our protolevn d tvtterest. vd florre!^ey it ,n*y'vnot loe strflicr uit4"n
$tystuons Jike : rnlfuat {s the 90 pavcertttle boatdiry Awne bwt {^di
tueshort q t*luysst
\Afit our
Oure vr,tr,rsb o\so wole +lruafi ?rol . R
cb"anqes l.uig s,.uyt dlrri ,rg4.1o
le;*},uve
ilsetf . t+ ig at $rnw ll.]<e-]rhue +t&at o" sfu,rdu,t rnagre{s taki
ry
an-,
ovrlrte cta,r,s v&lr'eY iw fi,a c:lo,t \ , so aA {p gee hia lrlexar
chrange dother rnYhil,,a te
!e*s mnolv€ on, *o rrruodelUn4 o,t*twupiiott"c
 6,2
D'Model & A,ssumphonr
\de sbaxt+o -t1y out ouy assurnqftqnr &
^"el3ust{X,#tetvvaLelt\,
Let Y derrerte *lne ro,ndsryr boo.nAtrlg trmfoft United (Wif-Un j W
Xi = Wfr i'tl,r obsevvakorr t* RzF b oanding metinod
Yt = i+h olosevvattan {"* W.itMA boaydtvtg vnelh,od.

n = f,irt b\^re sarvtp\e eif *.1,*ivate-rn"t(g number observalio.rg

+
P* h, RrF boarcling yrrethod .*t"'ne vt = -/z

m = LtraitecL sarruple sr4z. w e1-rivalenit6 ruun4ber ol>;zva\an*

-prt the W([MA boqy-dt .\"re
+
ry melh,od rn = E6

ie bodn" 4lte ro.ndom, /*ur*e vo+uablu -t b"th wueAh,ods (RzF&wiLMA)

"
dightbuh sn . thr, *u*pitow tw cLh,a{ i6 ,n,t
f :::t ?ndd^berause
Y"" uVx|i'arv rgndom vaxra,bre. -laker att iarur
1\'y:
0o -lo + @ .btL+, X ayrd Y oy€,
{r"rrr-
- mes t erLd *huV yLe!€y ,oJ.\uwt'
ne8.atfie value,s . Burt tn)€ ans assuvrar4g V^;ave./++rou* {lae ,**
witL k
a,rcoy luar* hLto( oto,tft +ive :,c axr* g s*yd ,
) d.ev wt"U. bo
aud& +tr&t "yobub'u\ y
x # brrg svytaLte^( *han
tnftvti+oswtaLly swLau.. "+ zLro
' n
r-'""t" at *he d,tst/ib-ufto'i-
+dwr^t
F(x(o).c o x-i ({,, ,)
Y -.il (1t,tL)
P tY(o) 3 o

x
0 o
2-
r
cap,uo[
Y, Yr, Ya f ave idepertAe*n
]
Bure
t{,
rno} $nt
 6,3
So the, f.*L rrvtodel,ting a,saur,,r,ptonr arle ds -$otfonx
Xt : tloo"rding -\ims + iil R2-r vnzlnod. +U?hk
Yt : Boordin6 {rme ill^ WILMA rnethod. {t^ALrt
*
X X.9 N (n, q")
Y ,9t N [-.9 , v,')
S e"act" ,twer
Ti".:o santplex &Ke inclepanc\enk

> 1u"u scte,v*i$c .antu^t'rt,

fvr,e questo4, L$e sba4ted ,"ri{n hlaa, dde tslhe bclte.t rrre,{hoc{ R2-f ar wi8Mene
+he tntoy t,\te rnn,qfiterwahcallttr $rrrr",^[ote tnjs quertton ,r^s.
{r,rd
ovt Arnercage sr Expected booxdilS *trnea bt* +1n% sayyr€ oy a^-e
hrue'thocl {a bz+*,ex 4taavr" srhe other

-(

4nis {s tlae sbafur quo ie 4t +t ir eorT:.et ot^t {o be trLte \'r,u+

. boLl4 ,rnz*hods ayig sarn-e. wlLvr* 16 a fu-ke,t rne]cr,od +f,\eJ.r
t? vto ott(ere nce irt-lneraattpl6 Ryf (as t{ r^n\l hov€ shotYu -hme)

p StattsHcal MocleL
We reca!,1 .fon+ e"atJ,s"el+ lWaV r,rrfi.at r"x. did s"qr.l.lltl- wqa ors ,k&otn*t,

7 t0e wawtad to es+\m&e U rcitfr" +lra

> A^ uJoc a,n es*Lvnato^r o) V z*ven w dorlA,?rD^f( Xt1y2,yg. --- -Xy,
-+hnt
7 tde^*iO"*u#,
"t f
i{ 1ou $"4 uru1uo +rrurrr^riu be unilue Pe+
accovding .{ ,ft
l {ne goal to o.s {,o come, cto.se 10 {rui R * usit't? oun daka-"
br,'"t |l.rlve t^9€, Itaw a dt{{oneutl sbty. Fc,rstty tr-6 iook at 'rshat oux r+,ro)et rt

Y i.s *wvaclexeA bv 4
E fP,i 0€o
)
uyt"knorac n3 L{r r le , 4,, %'
3ollw rn**a U
\, tar,lorgs {o O
f, : qondtdered
Any outcome covt be Fs : NitI be {hcjouni Pfirbabtti+h
q (xr Y)
drstriUuUi on pair O= R.X RX
(xry
ors po.tY
"f 1o,$1tbP)
) , As tcoV?t XrY lzrort, os'sunled +hAt
Alvqr- /
oire gousstavt X Al4ri Y qYz independeYtt
{r e&
,t1reR
{tu's, .is -
E= RX8, = RL
ie r-l
i"* [rtxt ty(tJ
prodJc+ 61 -thetroo PdtF q>e@r@)

6'e@r@)
 6"1

Rem embe>r -{fao questrrr h}e o J<rd €^Ark

x
x x ttseloeo ^,
x t* ,=
1, oT ,.tt > {,
{/
X
{x
r(
X* )l : ^,1
v x
. Fivgf tare wrust v'roke r,,rvhat ll)e afie dd^g
heve- 4w csv*rast -+o es+ima+i6vL
Tw e}rrde (uenn drqgrax{) islhz t^rd q +
atl probabtti+g a{stribu+ror4s a{ gaiv C},fl
perrametexi1ed b1 thz vec*tt

4l probabilit dtstrit2uscryr,r rl 4zt {, frf

9 = (r{,,
*+tnVe,e
for( gome YEasoyl, r^!g divtcle *he ve7Lon (ctrcLz) tnto +r&u^ paxJ(
ReW nrhare {r =.r{z-
@

h)so"vt.qald^a-:**aueshotLfUe',{owfuic-l'uYYJ#oYado%-lhq*rue
ir1 ,nh"at precrsef6
vroba,;l,it1- a#i U"tioru lazloyg- toe are yr.ot i*tawusbcA
9e ,ig (Sh;"r^,^lalthr. of 9.tt-attozt) ,u)4rt rr?',ic.lu :tr--gio* doea
Arol.
+* Vetrrlq +o.?

. Sgc-oyr,d, , *tta utloY rnp, {nqve. k'^ lated C'L(T u0e- axz wrb e*gw
1uesto.r,
enh; rtotvlir^q the. possrlat'tit5 {44At .t'< { 7- Soe*e*ullty
hfl^at l afi say irig is { knot"c e. c^n!+ betong tic t^e2tan' C.Tne

lueston u uyttetev [t be\onqs 1p Ao/(B

0vr,e wuqlnt ovq)re- 'J#. cnru+ratvLL tlourseL{
t"}i1t- +U{-..*l-f!1 foid*h'e
u**J., {^;[ ts ^fu'the ,fo squcczL woYr-'i^{ornt^att'crt
,1t * ytu" data'
fie "r,orre, your^,ecluce *ha^aof^ -|l,re r,o'L ' {o'n'rottcn' con be "1t^u"tJ
out + >^" dof,a"

f,T;ffi*ffi #f:*ffiu*l &? t 5[JfiJ #ilrr*"

b* fu rsseaychey ii
a"r*r{r.a to undexsta0 scientt$rc ordex soyale
u
caw be ?od aE
questtoru a Ygs or No
[rhenomenorL ^rhi.\,u
Does*lnis qene e$lecb -tlne disease ? Yes err No ? Gttectdato' 'findotrh
1

fu E coilpareto Placebo .co\tecfi{uta fin}o'rk

Doer this ieat"t,1e t€rrt ooqgh ?
Awe rtust nok ftrst cellrlh dota- &^zrt ,tot^ il^z hyPothlr^A
{rrrd,'s c\er,htj
Note *hot t/$e ave dot"g oru onaysi cn^ey tu:o som plw hznce- Wsts ajroo
sqm pte test
 (JrJ

> 4 f,"& l,ueuzisttc

gtateme^t tA {tt\o"ttry}
{ "l our prolotem
tevtstoru
I.* {'= rtz. or t^ {r > \z
8o 'f, sln^fi{ opproacl.u {lnjs problern f caru ov!t"'n"td
Srethe Ft.d"1
u3e kylor^ X ru ;\ (r{,, {r) and. t ,".r;i (1t, 62)
{ne somple meoruts good esttrvra}ort
T : L.L'N avrd. ts uYr}earsed
X^ ft,+ {r
; rf**-t"
YP : L, L.N and {'s unb arsed"

{ush go*-pore -the *t^co ie

Buk *lner^e 'ts a pro,blaru, heve . koe see +?Lis heto t,r)
lde wurst notsfrtget {l^at Xn ond Y,
ore nandsm vartiable.s -tlnawrselveA .

I Tne {tgure, {o *ina le{t shohrs tuhad the

1t**1, "&
R6 and f,. moy lok like fot
"! lo.vge, yvr and nrs f.t's c.teax(in *tne frgure ) '
{rat {r>{r But ourt sampte
Y,,
t x^ pr&zd.
So llT e
"p & *,a Y- st. xn ) Y-
tead fu afalecwrt^It
l,re.uylshc abong wi\l

fi, uetter Jo inrrtstgate is -lo irntroducc, b"l,l et\<xvn {oveacln sample

b"tf,*^^y "
witt depe,nd srn" {ae !^o { observa;hnns ov"rd -ttte vanoanc€A

X" Bu{,{er (vt, { )

{, .1,
Aeralrnds 61tu r-\, De?ends o7t rn
o.r{"a 51 and 6.

{* Iqf^ cnvpar,atizr :*, the yavrdrnt $fuatr^ahra,e het&J {o f^

irn

6r^ct % . we expect +t",is Pufler *o go-ho 5evo(o) qa T3Z

t":F.l,r o arlwivaler,* {o sEinft *ln-e w.lt
wtl !o'to dofi q"a Xn "t proba)oru-q d.tshibuho.rnr
& \^ _y_=* + and, +1^o,/L *tr-e
^# {,
)€gull hle Xet heitl be much, mure nobusL
 6.6
tr Wattirr.++irfte t^" E.R,.
-If,e next firobtem *hal unre n:ant to {orvnutabe lA orbouh .ToHyttrn-ex on cr$ erqge
ivt the e,h"rgenq .ra1-m-, +lene- tnle dorr't v\toYfy abor,tf -the didtribulioy\ olrrt
li, Xr . , .. .Xn ,(ol,ton: . we aHe ( trr, +his prolotern) itruterested tn horc rarc witt
ir,tws*tgahn4 it
{orvnulabe -Hne pvobterm, & hon: we 8o on
Ao tlneye/s o. hospitaL calla{ 9rt"cekon- ?,l,orrrrbrso . ?aop'e >oy +h.at i,ti
o- horribte ?uosp,ttaL arrd gou h-o-ue lro r^:alt m6rc
-tir.,on r*ua.[ {ov +he
Evnex4ene4 e"oor,- (ER) trenl,vrtehll .
5o t/$e collecu n santptea ,indap enauX\
Xr, Xr,X3 .xn be-,tomXt^?+o Sonns distzibuftot E[xtJ = rt
Tv.,*. , iru c,or.utraat to tlrr. boartdir43 problern, on!
tr*t abse,vv*h6w AA v\,pvetla"i,s
one somp[e ts collected So is a, c;-1r- *ample te* . €or cne *rnp),e-
test L$e vaeed benchvnarkg te l+xed trtumbers {1^4t I eArLbe comparned
i,o . We go {n *lre CDe roeb,rtte o}^d +ann" +lr1rt? tl,re arrernaqe l.itvwt- o\.er(
ui[iona ][ dcta,lolase-vva,turna tg tso riinutea hznc.e-.this is?ru2 benchyrra,nJc
Wg sce \^-ow \he errtluir'1 Lorv bz dole
Iv@ Wos6tet l,,cltavqe , LaG cotL h.q Lrso, Ud-d* b ex'Lrr'n+e,|4 cauhout
qbout e* r*p"t^t[ 4 hzv hosrpital . f"o sheo ftrrrt u\a(ez
,
1g
L
at X" + tu{,[ern( 3o *tr'c,w .l1 < bo
--
defend"r mt vocriance o{ xi
qnd tn-e. s^/yl"pf e Si6e

*tn+Jazv pur3oyr, Xdi colt

Wrw Dr. ,t{crtse iL ,tL6V r.,ruterrested in, 41,^'u

Vi"s{-tAi ra'3'tt a*Lart . 5o h,e '(orrrt-r.,tLa-fiA

t ! >Zo
{1x || -htJfiern > Zo +lrvrL 4., Zo
#
d.f*da srr vaxdayrce o{X{
and -[ne saw,plc suu

up unlil hord hre hnlre on! sdstect the pvoblem o.ncl

,rrot[, it dorzsrt mathemafug
vua r^rtt[ T[ot^] rno/e. on ond
do some prectse co[cu\attsn's
 6.7
P Ex.ornple I on 4q7L t'on d crmty somp led Avrreric,oru ) zL'/,
suweg wos c^nducLad- ln z-ot7
f!report
r,o ge;g ab \east some ol {tnej,r neh)s orL Ymttrbe . Con we concLude thok or
rnLt 1fijl.J d, Arnericona 6eL atteost scrme c| {l"eiv neh)s on lfo,rbube ?
^
So *lrue rnodet u ar {o[or^s.!.
Xt, X:, Xa .-. -. ..Xn
".., Ber(p) ;
*r = 4q7L
5(ornc this p oa o{- nonr coutd}ae avr,y number< e (orf)
Bg Cerrlra,L J&uit *lneoverrr (c.ur1 hle cctn sqy -ttr-aE
)G
JN
P
Flo+
Y1->o
i ro, r1
Jp(r-p)
Ti," repor| sogr E = ,aZ
-lfie to enquire ^,aa : Do atmost != t.==) Americavrs ge.b sorlae
v.re uoanbed
lueskon
,"^" from Voutuloe. To thrs problern
^lott
rr,rc Qlo {s Puf ? = L/g
"+uofy
tr.i+tot! p coutd holt been a^J vaLue bro o ond ! hui {or+ ot},t sge<.f"
\^)e. soy leb p = "33 (at 1/z) tce {kre -hrutlo .& then c-olc-utoFe
"^4,^i,
X^ ( ,3L - '33)
JT P 4qY1
.38 (r-.a3)
= -1.4qq1 = -1.5
Jptr- pi

Now v.le lqvroul p likeo- ito,t) Varuabla (sa1E)

fi I, - - loehaves
p(t - p)
^-1fl"\r!"fuoLk.

&o *t *1*tI a)

,lA
fl^** llt'al {ruz vatue { P = "j3 .*to^r hL% o* 1
b geh a, re.ali6ahoru * Z =-1"5 .€or'{lrat Aset- the dtstribution
)i
t0,11
4nd 'l |ttA 4^e E rs ved;yA belwcg,-
- k +io 7, \"9t*," Probahititg q57.
So Y = -1,5 'k a, pouibte. reolizarlr'on
if I {ar<< q57, prd"alarLt\ *
bznchvnatk-
^y
7'/.
-T -2r -fo 6 2t 2t E(
-z -z -Io I A 3

lvri,s is +he bo,t,rc puvtcilate : Yotrhcrre a \,\rolhe^;^ . As.surrte itg +nre ?ltVin,daka"
+o f^a tf +tae valua attarvted .rs a p*sltr" re,a\iz-aUa4.
 6,9
FExorrtple Z
A, cocvt ts lnssed Bo drnea, Heods are obtaitpA I3 4tvra"ea . Can hle covl cU"td,o
tjaat the cnuT- tzS srg.rntftcarr.,tt6 tlvufarv l
f,ust Ji.ke *he prevfous exomple rr.re sotw /dnluire heve
Xr, Xr, Xa - Xrn,
^, Bertpl
Ttat soqo1.^ cottecfed h= Bo
C LT *e.\\s us

ln
Xn-P t-#
(d)
,ti (o, r)
h -+oo
p{t-p)
\" reperrt sqy (or suyr^e! s%) x,, lb
bo = .AZW =,42
uP!"rkt v\or^r hre sarcl p e (o,t) rn orcre.t {o test otrr hg{rotheux
r^'e httl q*,,^e beforeha,d th.attne .-i,-
ar {aiv -rs p = . 5

j,. X^P .43 - .5

gtr-p) J6 .5(r-.8)
An - r\n
,l

Noro *t autlun4 becnmu : &ufp{aogs Iot*umz 3flir,r, .,,,t

elctta procluce q- r'r-r-urtbes+ -firtct cay,t" Vte a vafid
toi, ' Doea m
1
reali a
[trrmal" "randnrw valrrh]:{e
X.
trx+tow fi s+o^d.o.rd

Tue onsNer
%
oiyr & Yts X alArrh4rr val,ue be{r^xm -L *o L tot+t4-

?rob*,U{3 9s'/. hence X = _.77 rs ptaust ble

Corluhtuo\Z Tt i s not until,rJy il,1^\ tne u:tvv /s
{*f"
Note : What we did here, we assumed that p = .5, then we looked for contradiction in our data
to say otherwise. We were unable to find any contradictions, but this does not mean our initial
assumption was correct. This only means we are unable to reject it with the data we had.

To understand this more thoroughly, lets see a weird example. Suppose I hypothesized that I
am a superhuman, in order to reject this hypothesis I devised a test. I will eat a whole
watermelon in one sitting. If I fail to do so my hypothesis will be rejected. I sat down and ate it
all. Can I now claim that the hypothesis is true and I am indeed a superhero. No, I can just claim
that I couldn't find any contradiction to my test, hence I am unable to reject my initial hypothesis.
This does not mean its true. It just means my test wasn't able to reject it
 6.1
F Stabtg+tcaf Formulatoru ,J- ryprtt't sis Gs+ireq

hnd a stcttsHrat rnod.eL (E, (Fu)e€O)

fr,r.is is done tn orde.s *o
ensuhe {hot o.lt tes*s {trat h}e rr.xtl do {*tho
do not fatl inlo d.{f<tetrt cotegortes Like Z test or t- test. We r"r"nt +.
Iook qt qlt tests atozrce uryrde-rt a wti$ed theor6 & 't ri" $rrvutatcirw leK
me do tt^at
> TA undershand l^vhat nre elo next, nre-wrusf {.v"t recotl r^Jhc*l it -"*rt t t
a vnodet t ba t"p,ltqf.it*\ and rnrhat it
meana t6l( U .to be tdenfi$'abte
Tnr:s +8"r=r,.ros drar,wt e61t'ttef
t Ft]".o Ne ossumed lhcrt {heve hhs o. B tt"at qenerffei our
dota. ovtd d{ the model fs l"re[ speerSed then
thts P wttl lre ingide tlne crrate r'^rhich rs
tlae subset tFoSReO
Mathematu^alty Tne vno del ( 8 (q)
) is n:etl-
, se o
speo$ea +fr -ti"e.rne cxJsG e* e O st Pn* = F

Att probdbiti+U distribu+ic:y\A

r^cher€. P is tne distribu{tcrr {raL ga,rerofed d.ato

\^)e hqd ideniii0iabiu-t y fut ro5vnan terrils

Dnvra.t{iar[t +
f,n o' uruq qe valre rl g &n- O 'there rrrus+ be a unique B tn" tft3s6g
u+0'€,B+8,
$o ruor,t) "t h)€. horne o. tnre,tt specrfied rnodel ay,rot I rh, ldenh{r.oble hre car\
vie w t[L3".o e4uivaten\ as +ha. set O itsetf
f Pelee o
o
0r

g*
o/

{ne is h}etl spedfiect & I ts tdenk$tobte 20 thsre- .ls o, u} r.1ue I in O

modet
sornespondrnX *o e^cW -Pe i" tPs ?eeO UVelt spe-c-r"ftra*ton insuw,s
Jhqt dqta & ganerated frorn one o$ +r,u e.le.mer* "1 tPeieeo & icler.+t$'a,bilili
ev'\sures ,hle con $"d_(equtvalentlr) .l,rne gthat
Eenevated *\e dofro. iro
oU, parametav se* O
9o !^'9\et r,n.l€ clo ts c\ivtde
O ira,to *rno drsjoUtt par&
Let Oo and Or be dtsi"int rubseG d O
No6 {^ot theseo.rne dr=jotvtt 4o they do'r* share qy c,crnrmsn €/
ThEJ con h oY €r ne;d not be iottectruey exhals+ttne
 6,1a
ie tt mqy so hoppen -thqt ffi- oo u Or = O but tls nob hecea&Gy
Y
{owevet Oo o O, = fr (nqlt) Sis is mecer,6oy#

e+ t ---+
+akegutosc.(
accordtvrg {o
lFe3eeo {he probtewu
o o

)tglpotheus is o ques{tc'rt *hot ceryx orst< tur, Etr*us\a" ,like

. fe the cotn -[air . xs ^rc
wiLMA_ betteY tlnan R2F
. fore wot+1nq *rlrner tn ?ytrrcetran glointsbrao ryrx-{hnn usuo't,
. {s the- ablil b?XRr tlron Frl,acebo . Do peop\e ho.rre cr" pre{erence in hrsstng
I^le ,rua{n"ernc^hca\$ fo"*u\otutt?,m, tn tevnu tJ g{.u( povameLal d
9o ',,rr c.onstder luco hYpotnefeA
Null hypothorie )-l.o : I G
r\t*ernatetsFdth,i^ 4 : ge Or

:i++ *y$ tmpovtant, to ,rndevstond, uvhat ,,Je o* ddryf &th"e *erntinotqy tho* hle \lae
}{o and ]h ore vtok syrnmektc . J{o is usua\ reloted {o stu}uE quo . Ie uruless pr5\ren
othern,&e , *rhl r^toutd hre sqy c. c.oivr ts %aised tr'6n'1+he beg,tnruing. Or a person ts
fnnocevtt -r.ryrttL pro\€n guiry. Or a. cWug.lan^s the some e$ed aA cL ptacek:o . Orc
\/\p, *tme inER ts svnatleythan equo&to 30. Ortoot?u methods' WitMftE RzF
^"t+tB
l'rarne om a\r€Yaqe the son .e trng .

We *est flo against flr . G ss stahlttohl4a its our job (r^xlhlhe hetp{d.qfa) {o-lm*<
fir evtdence against Ho . rde nnag t-rY,rtsr reject tto or{- mo1 VJhal be unabte +o
1^rc wttlsly
roject flo I but tlire & uufltt ne6r sag *Hat rnrc accepb Ho .
ig ,yhe,"testtry Ho againstr *{r rnre tn}sre uvrabte ! f."a.T*{h e,vddenceto
a"j** Ho, henea h)e atl€ unabte to reiecl tto . {,us effi does not mea?lhrs
fJuew tt (Ro) -to be true. we j*t wernettt o.tole *o yejeat tt
fie dcrto is orr\ used to dispror,e Ho tw {avout d H'
fu, E iw -the Ee Woittrrq ycrrnn hle con [and d.id qatlier) {orm hypollne,rrr irr,tr,'.:o \"]qyA
we should choose hyboth.rr^ @
H" : t>so Ho '. J.t( 30
AS H". ctss\rmes innocernc.e
Ht : -tl< 30
)ir : ,,{ >30
p.rorcn, olhey^x.se ie
uraJrt{,

@ it soys Pvinceton P-tatnsbvoo's }oahn6-

trrv're is jusl Jrke ofr ol'lrey hopihtl
 6,tt
[T:est Stafistc
So the ln:ty trp harre defrned our procedurne ,ts al {ottot"x
-) h)e ho\€ ourc olpsevvatorrs Xr I Xz,X3 . --.--.Xn sll t{,cL
-+Tzrc' is o' stqtishccLt rnodel ( E, tPo3oeo)
& {here is a dtsbibu+i'nP qercrattqdf}a
-) is rnetl specif,ed so h*' cctv't clo oll otrr tnveskga{ton
g
o.r ba"ura sl d
* tnle wor.r.t *o ask questtonl , * $rrnulate thei
mathernaticatly
) 3o r"re pvopose Bp"th*es bosed
poseo sn on rhe'
*he. subse+s ql povame{er set
o
-> Ho : G oo
g
Hr : RG Or
make a. Junchort co.tled test e{atrshc U
Nor,tr \^!e
I ts o stahshc te o measuvaHe functtart ol-*oy
ctata" - Ayrd nle yvrake i{ speci}colly
Lo g*r,.e rrs {tna result d *r tets+. fn" h}e $vrvtulater:l otrr probtern,ir
can "ornly haw *,,00 sotuttortg n, stth.e,v Ho is rejected or .il3 vtot rejecLed
So 9 is .Ju,ne{tot+ lhat q.r.\ es cK, ottthtt on$ too valler . Tt* best *o *rfl^
thege -tuco valuer os I and O . 8o
A, tQEt ts o- stattsttc qL to, t l guch {^rub

u 9= D -{.to ts n of ve-)ecteA
1t V= t ) Ho ia ue.je&nA

Now V is i+set1 an tvr.dicq+orJunc{tsru 1l- , bec.q.rrse it *o,kc input & giws out l iro
on,.d'
g = ttr*=.:ffi,k4*tfl^ I
the t'elttry pvoced-rrre rnrttt be
aceording {o {he na*ure C +." tegt an-d the
y<tnd { vandom vaxr'a}o(er avrd &.te1uali{tea tnstde bu* th-rs i=the general rdea-
) Lels see {}ue codn exomple aqoin-
\Ne wovtted {o see if cotn- wqs uyr)oatsed trr ncrk
Ho : p = 172 , StttU 9uo, Cotvrshoutd be unbarsed/fnnortnt be.foveany ccnnyetevidencz

Xr t p * t/z I fne e.LtevnatB hypaneril aga(nrt uoh*ar Ho & tz,sted

L.o gee rnrhot r,ue alid h.ovp. o,cco,ndtng tr ouy {orrvtu\atian
O = Porometer set \No.s (o, 1) tor +na parameter p (here p)
lvt order {o enqu ire rJs coin baised ?' rr.re pa,r{ttconed O

o r>O t W e asked S€Oo E P=Llz :Ho

t, 9€Or = p *\t2 :Hr
o,
S(orc r^re rnoue *o the tes+ s{atisttc . tPis as{qtisttc og_it does
catcuta{ion otr nn (ie da}a)

Y i. l"m \ In.5(l-'5-.5) >C fot sorne c. > o 'tP

hlE-l
X\nesootl q.BiE arr inlicahrr
I >c isthe *esttng
'J's(r-.s)' prooiluve
 6,12
P Errur,a
tsthe c i',r- previous V'lHovo ts t+ eleerclec\? t"ce will see now
Wh.ot
&o os a SYraat philosoplrr-r once said
" \r)h"r" atest wtakes o. dectsiovrrtn"per$rrnance shouldbe measured q iK |.:Yr6{
*?4-g
So rarc hstll de{tne sbrne nerJ terrn8,
. Rejoctron Re.q,tan
"l
test , Rq ty
Ag {ne norne sugXesb , its *he regr6n- yuhere test g mtlt vejact ie nutl \y p6\hotr^
}'!,0
B'rt hrh.et elo hle vneovr by regron 4{r t nptrr+qttt +o Vyt&<,th4t
I 9 rcas a. (an} so.J
pbahlttc - $ to.k€s i^ dala i.e samp\e X11X:.rX3- ---..X,r- . Trte samptes +hat
reject *lne hgpotl^e^^A Ro betong +o +1.," rejection *ff*,
RV .

&v = tx e 8".. Y(:r) = tt

r^.x, |4rrow y {s ayt ivtd[o"tov 9o
\y L lx eRv)
a
Lrrcr Tybe I & Iytbe 2-
'Tb undevstand a,t *he fotlowtng toht*,
the +r,uo llres of evvore rnrc look
esL
Ho Hr So !ca[+k ts llo avtcl *he' teat rejetls
w,t'uevr,
it
u4vror
it is a tgpe I
rc.ahhy ,r^s H r qrtd tha tert reiect it
Ho 'lyb"t Wlneyr,
'tU
'is htpe C- o'ror
Hl t/pe\ "

H' c'onbetrue
Atso notz. {hat we oresag,m6 tl^at ''o,.\ib. .^ ho !v.z h*'hoth%ia
{catn be- }aiv . i lrir"ot*r', bu"irlrlocent . A, a""t c'o-rtYu *f ^;-l<a'
a ptocebo 4ne wa) hla
jn4'"ua crur.lornu'tt$sr'-r". compella u^'{n ejt*ter
reject it olr woh fr.i".+ it.',6^ +t4i^ da- oltsUu,rXurshtoru one shoul-d kesP t,^,
{ne. back o+ ilrlL t^ i'"\4.
TvpE L evror
\yt?u L urrar dn o t*st { ,a 0J t"*d*
o(v ' eo '+K
Type 2 eT rsr
Tspe z evrar { atest tI t* a {uc*to'u
?v Ot R
s ,-+ P" [V= ol
 6.13
But i{lis {s ?tot just 0. h^afi,or vrralhe,mahcltL comprutcahon
$ r ldf LQ,e hpn:
twe-{rrwc,ttons dqr & ?W axz c{e,frned
For eS.Lstg took o.b c{q, : T5'lae I ertur
By d.ff rnre knoN +l^at i{* -l{ae evr* o} ve.lechrtg hgpolhorir Ho nvhan it
Nqs tytdeerL true .
\r.ylut does it me-an {ort lt" {o be lrue ] rdett accovcti rrn *o orrv lorrnrilahor
s
t'1, vrteons g (fe the 'neaL PArcameberc) betongs to Oo
t$h"ew witt -tlne *est reject Ylo? t"lhen *he s+ahsHc +hof t^)s ca\c-r.rlaYs
wtth the heLp d ctota, grna \P = 1
8o !^,he cons4r t,tcL a pilcttctru I
o(q : fi [V=t] e€Oo
V 4r-)
1ni" Sovs i^= s op H' is aclua\ *ru a
Jt,t* g t's &st +olcl rlt tt reln L+eA
mopfd forn get Qo

ldr^rUqxlti- LQE l-ook at pV 'w L e.vrw

Bg de-firw \^)e kvroto +l^ot its ff1,,e, zvYo-r wrywV rejerhng }to t"rhew i'vt'
*
o,ch,tolitX. it l^las vtot trus
l^l,uat doeg i+ vreon, {ov Ko {o not be true ! tnlett accoY dinn- 1o our(
-prna.r"rlattor,''i.f means' g [ie {he red pramebov) Welony io Or
When r^lrtt *Itz hegl bz trnoble {o vuluLL H, ? lrJhe,rn, {lne Atcttsh:c tlarf, we
Calcr,rtate r^Fith the t rlh t& doto" ,[r*t Y = O
Lo hle cor\sbutt a tme+toYr,
'o

?
v'P*tv=o]
4
UeOr
W
.tvus s&ys our t.h& s*6 4{o ts vw\ *rue
lrtbIIs tasfu b{d
us {hat Ho
,Wud pom canno? be vejecleA

ht* Or

JHese orlc o{-'{h e vno>L *,r,f," de4r'r"i#ovr* 6ne, uvt cane fl t^ri+l" 'hteltia
^t<-e-
?" v.)er 6{ a te*t V

-fi
Y
i"{ (r - pv te))
0e Or
 6 .71
Noto r^M closely loolc at tae de{trr gl poh}ey Nofe tne orgurnent
ptt E F* (Q= o) &obob,rr.i! +fir* -the *est t's unqblelo reject Ho erre,n rruherr,
= in reol.i\ )'b'ts noL*rue ie O
Qiuen o € Or €, Or

\^Ih.at rcou\d 1- pv meeh, ?

f - p*
= G" profcoUi\i+y + vejeedng Ho rn:hen Ho {s noL trueie I e Or
= IP" (V=t) qiwn B e Or (ie ngectSo tthen yotr reolly shnild)
(nu ,nore {t^is wtll be , fnu vrlsl€ cern$devrce !^}e wit\ have tvt our
p+nstoab'rUtt1
&s't . Ie h)outd pe obte {o sqy o uY test reierls R" uuhen Ro is nok true
',nre
\^Itjh probob,ati+g
J,- p*) But
. tr\Dd lt^Gt
P\t,
tli ^of *+*ed . Y witl take,
dt{feve,vrt value,l F" d.r{,lerergg B,s tn Or.
Br.r-t c^. churrn & only es ghcrn o as iK r"renkest .!Jnk . So &.rc aI[ v al,'u.gr
-l
s+ S € Or VOe calculoC[e.1-Pv Frlv=r) avr.d {t,^d iK r*ii',rrr,r,nn/ in$muvrzr'
=
{er"tz
fiv= t"[ (I - pq,0))
Ec Qr
1rus ts the irafirifir,ne t"e t"* powrer + o' test
> Le*st/ b undevstqnd rthat ts hoppnirg hene a likte bzftelL
Suppose hle desdgtn. a test V . qnd rnre wor[to 'oloserve how tr, (qr= 1) changes
hlt{h I ' Jfus ts. a SenlaL pvobatcit\ as we ayr, not ctairrtilg th6t g e Oo cn Or
we wo^t 10 ona\yse t+ d
fr the yuhote povomefev set O
fn srder to visuoLlze hle witt lakethe ER wcritt4g aime exo.mple .1^"tybdh.*6hrere
rto: { < 30
)tt:1)3o
As {>o bU 'tln" probtem g*atetvre-nt
hs(. kvtoul
O = (o,6) , O"= (o,=ol , Ol=froro)
Nor,c if
hr€ hove desi6ned a clescent test V . Tt ghould be {aivlg crrrttvruoua in
t" ^.t*rt}t l*o + ory s € O . {tso Fo (v=1) ehduteJ be crose to -z.e-yo
ushen d e Oo i.e i+ shoia *leq
Shoutd be close -lo t d*t io lnrheru Ho is true. Atgo Po,(V=I)
w&ren I 6 A iu it shoutd rutgn" hr'gh probabitfy rejaeb
tlo when Ho is not trua tne gr"gl" shoutd looL {,rke *{ais:-
F. [Y= t)
Nor,r: note -fl.roi ivt re6,t ovr e € Or
4hz -Fs [v = 1) obtatn s nvuin-irvtuwu value
(ie in firvruhru) ab 'r1le = 3o
fr" eln > Bo is quivalevtt to
I - p,f, (raf &{Yr )
$e+tn ir tn{irnurn ts Utdeed*he pourer Tf*
g ciq,

-r,l
oo =---5; ^xrs
 6"3,5
p ueuel x
Nor,r: thab vnrc terma S de{rtar*t*ovts Jels itnal\$ ansh}ev
dedtnecl sOme, or,uc qt erto,.r'

$ houc *lre c iuot deciclecl ("e-e*bs,, lhe coin{oss exomple)

Le*srec,all, cte(0) = IP*(W=f1 9€Oo
c(q h)as retatec{ b Ho vr:*ren it rn)qs ir"deed *tue '
tP" 1 e-n*ox, ie neleol'irrg
an e nno.in P,abtevla &
4t * dtsclssec{ eaxLier il,o re.\ated to {'lne sta+us o[,^o.
botLr -tglr" t ond \P" Z eYrors can't be 14 i^in^izsd gtmuttaneoudl fo
underslant +.l^r:s lets see on exantple
Suppose oL peysort {s ,atandinp lvt acouvt {ox a cnmplairt a6ar'art Wmlnex
)to : Twe person {s tnnocenV
,t r : Tne persorrfs 2uith;r
Noro l.l,o is the ste^tus t*o q.s h!€- la.vTow q^U person ts innocen? untiL Frc^ €14,
u
B*L+l q cvirne. Lets
see r,^:hal Nne ervor hloadd l.odt< Like.

v pe t ;
--i-r,re
couy| dectores person qurt\ , t^shen ivr, rneaLld- he drd v.olr
Co mmil, any cvivne.

2 : Tl'te courtt sets *he Person f"a" , tthetr, ,lr,r, realih& t4u drd, comrnib
W" ttve gliv:y.,e.

Nor,u osk youvself rn:hat should {he cruv*s do tut orde-rc vternert 1o cornrvrit'
tFu f evrx ! . t1 .-that t^ror.r4- noonr. witt b4
Trre ovxsrd er i; s trrr ple LeL sverg o dy 30
b
lLl" dL -*te c^ir'^i^4lA
coyrvicbed nrroql,y'" {sre\ner,c this stratevX,r6 lnJttL

Ao ffe too. the couvk do tn czrCer neuer'clu cornrnit

Nor^t ask yauvsetf \,ol4a,l shouH
Q.1be
2 !
erwr
Tvre onsruOr ts sirvtpfe . ?ut o\rery on€, iw iait , frr".
il^o* No:{
.,,^o
r.J
crtnruiytal corrld. .schpe faus* tnrg. o.rtt'144. But tlnis slratevqty rnritl -leadb
oLL tlrre- lrrnocent peopte ( tnrho have" a veport ^tNwL {h.-)
b".^{ ended ug
iw ioit
So tnre co.r,Ut *tn;eok oyte kivd of, errat nvtthout
tae oVner. afJ€.+i%

go r,nyhat rnre do rs put c^ tareshold * aF. I oyror . Ne sry maybe

;il l./ilt be cerses rthevr tlne o'.ctutaf r,vtrvdrn"al \^!itt nralk J*
u tratt (re!P.Z zyycr). a r".rtll unt oLLorc +n1 lrrocecture 4o anvi&
4tn -thavr.
d oA'o[ 1in e iwvr"oczrtL peop - $a
1

vtiote Le

A )zgt, l^^ct; a. tevel oc I+

dv[o) Ld." ,d gcOo
 6. 16
8o barical,tg d*^g t{nis nxtt lce a. hord ld, ,t " S"itZ aomp\er Sorns€.hfk oboW

I *est V 'l^aa aln'rngtottc x tl.e-,nnsL 't

)irw oc'y"(e) Ld , r(ge)o

rL->@

> t'"sU Re v isitecl

vUe eard1e.l1 desodbd V e the *esL $uvtckcrn as qvl tncttc.o.tor
f

r.y Lt Saryre calcu[akon ow-!( daYa, !

=
-iru* some calculado, v.:tl\ ntt"" be iutelu"tt\
+ $rr,r" atke
aru
r4
Here Tn wtil be c^Ilcd
11t" tert stctrsbrc aA
lT. l>c- i6 a ca.lcrrlatm.r o*{f.u do.ha.
lTnl < c For eX.. tr. cotr,r zuample
ii-p ar *ho T"
Jt..,n d*'n "6 ted stxhshc ri ,flexibte JTI
ptr- p)
h)

p one €dded vs -[oo stcled *est

zilhet s a
oyre srded*est ,
ffT^T:f;"T.;.,"S.il11.,}if^:i^rfitgi
Ho:9(0s Hri0)0o

Two &ided test 3 He.rre. r^re- clnfinz I 1ro a speci{lt Oo iw *te H, kufrot{tetU
Ho:0=Do Hr:&#go
Ue{s 6ee Vtor'n: -dnese det'n i{
3 lrt'ed +o tes+ if roilwra bettq : On e sr"ded "t{nzr ),r(wfunrA
1, Booxc{i.^g Mt t?rod'

t,. Wai+ir,6 fiml- in ER e -(3o Ht : 'r{z3o " Onestded'laslf

Ho:rt

5. Kr:ss exornple : Ho" p = '5o ;

Ht P #'Si3 " Tino sided +esb

t. Eo.i" -t,n ! {1, p =.5o , Hr p +'5o : Tuoo sided terk