A’

LEVELGEOGRAPHY

GEO-
STATI
STI
CALANALYSI
SANDPRESENTATI
ON

Pr
epar
edf
orMEGTA2017Annual
Conf
erence

ByAmosTendai
Munzar
a(0773245970)

1.
0Lev
elsofMeasur
ement

Measur sdoneonv
ementi ari
abl
es.Av
ari
abl
eisachar
act
eri
sti
cunderst
udywhi
ch
assumesdi
ff
erentv
aluesf
ordi
ff
erentel
ement
sofapopul
ati
onorsampl
e.

Thel
evel
sofmeasur
ementar
e:

 nomi
nal
dat
a,consi
stssol
elyofnamesorl
abel
se.
gfar
mingr
egi
ons,
cl
assi
fi
cat
ionofset
tl
ement
srur
alorur
ban.

 or
dinal
dat
a,consi
stsofobser
vat
ionst
hatar
erankedi
nter
msofsi
zeor
i
mpor
tancee.
gcl
assi
fi
cat
ionofset
tl
ement
sasA1orA2;
orashi
gh,
medi
um
orl
owdensi
tysubur
bs.Magni
tudesofdi
ff
erencesbet
weenr
anksar
enot
cl
earcut
.

 i
nter
val
dat
a,measur
edonascal
ethatdoesnothav
eameani
ngf
ulzer
o;t
he
zer
oisassi
gnedar
bit
rar
il
yhencer
ati
osofi
nter
val
scal
eval
ueshav
eno
meani
nge.
gtemper
atur
e.At
emper
at eof0oCdoesnoti
ur mpl
ytheabsence
ofheat
.At
emper
at eof60oCi
ur snott
wiceashotasat
emper
at eof30oC.
ur

 r
ati
odat
a,hasameani
ngf
ulzer
oandt
her
ati
osofmeasur
ement
sar
e
consi
stenti
rr
espect
iveofuni
tsofmeasur
ement
.Theabsol
utezer
ois
uni
quel
ydef
ined.Forexampl
e,i
fthedi
stancebet
weent
wopoi
ntsi
s0cm,
it
cl
ear
lyi
mpl
i
est
het
wopoi
ntsar
enotsepar
atedf
rom eachot
her
.Adi
stance
of20km i
stwi
ceasl
ongasadi
stanceof10km.Ot
hergoodexampl
esar
e
amountofr
ainf
all
,popul
ati
ondat
a,mi
grat
iondat
a.

 cy
cli
cdat
a,dat
aconsi
sti
ngofdi
rect
ionsort
imesi
nwhi
cht
hemeasur
ement
scal
eiscy
cli
ce.
gaf
ter3590comes00,
aft
er31Decembercomes1Januar
y.

Rangeofst
ati
sti
cal
anal
ysespossi
bleoneachdat
aty
pe:

Page1of1
Nomi
nal
andor
dinal
dat
a;ar
it
hmet
icoper
ati
onsof+,
-,xand/notpossi
ble

I
nter
val
andr
ati
o;ar
it
hmet
icoper
ati
onsof+,
-,xand/andal
someasur
esofcent
ral
t
endencyanddi
sper
sionar
epossi
bleandmeani
ngf
ul.

2.
0Uni
var
iat
eSt
ati
sti
cs

Uni
var
iat
eSt
ati
sti
csr
efer
stost
ati
sti
csper
tai
ningt
oonev
ari
abl
e.

2.
1Measur
esofcent
ral
tendency

Thesear
eav
erages(
mean,
mode,
medi
an)whi
chl
ocat
ethecent
reofadat
asetand
ar
eknowncol
l
ect
ivel
yasmeasur
esofcent
ral
tendency
.

Themodeofadataseti
sthatval
uewi
tht
hehi
ghestf
requency
,thati
s,t
hev
aluet
hat
occurt
hegr
eat
estnumberofti
mes.

Themediani
st hemi
ddleobser
vat
ioninanorder
eddataset.Thedataisusual
ly
order
edbyar
rangi
ngv
aluesi
nascendi
ngorder
.Lettheor
der
edv al
uesofadataset
be , , …, wher
e i
sthenumberofobser
vat
ions.Themedi
ani
sgi
venby

a. i
f i
sodd

b. i
f i
sev
en.

Togett
hemeanofy
ouhav
etof
indt
hesum ofal
ltheobser
vat
ionsandt
hendi
vi
de
t
hesum byt
henumberofobser
vat
ions.

Thepopul
ati
onmeanofasetofNmeasur
ement
sisgi
venbyt
hef
ormul
a:

wher
e ar
etheobser
vat
ions, i
sthepopul
ati
onsi
ze.

Thesy
mbol i
scal
l
edsi
gmaandi
tmeansadd.

Page2of2
Thesampl
emeani
sobt
ainedusi
ngt
hef
ormul
a:

wher
e ar
etheobser
vat
ionsandni
sthesampl
esi
ze.

Exampl
e1

Theamountofr
ainf
all(
inmm)r
ecor
dedat15di
ff
erentl
ocat
ionsi
nthecount
ryona
par
ti
cul
ardayar
e:

44 5935414625476054463546413419

Fi
ndt
he(
a)mode,
(b)medi
anand(
c)meanoft
heamountofr
ainf
all
recei
vedont
hat
par
ti
cul
arday
.

Sol
uti
on1

(
a)Mode=46mm

(
b)Medi
an

Rear
rangi
ng:
192534353541414446464647545960

Si
ncen=15i
sodd,
medi
ancor
respondst
o

(
c)mean=

=42.
13333333

=42.
13mm

Achoi
ceof
tenhast
obemadeonwhi
chmeasur
eofl
ocat
iont
ousebet
weent
he
mode,medi
anandmean.Whenchoosi
ngasui
tabl
eav
eraget
ouse,t
hef
oll
owi
ng

Page3of3
f
act
orsar
econsi
der
ed:

 Thet
ypeofdat
abei
ngdeal
twi
th.Themodei
ssui
tabl
ewheny
ouar
edeal
i
ng
wi
thdi
scr
etedat
amadeupofar
elat
ivel
yfew di
ff
erentv
alues.I
tisnot
sui
tabl
eforcont
inuousdat
a

 Theshapeoft
hedi
str
ibut
ion.Themeani
saf
fect
edbyout
li
ers(
ext
remel
yhi
gh
orl
ow v
alues)whi
l
ethemedi
ani
snotaf
fect
ed.Themedi
ani
sther
efor
e
appr
opr
iat
eify
ouhav
easkeweddi
str
ibut
ionwhi
l
stt
hemeani
ssui
tabl
efora
near
-sy
mmet
ri
cal
dist
ri
but
ion.

 Usef
urt
herst
ati
sti
calanal
ysi
s.Themeanhasamor
eext
ensi
ver
olewi
thi
n
St
ati
sti
cscompar
edt
othemedi
anandt
hemode,
andt
her
efor
eiti
spr
efer
red
wher
efur
therst
ati
sti
calanal
ysi
sisneedede.
gconf
idencei
nter
valest
imat
ion
andhy
pot
hesi
stest
ing.

2.
2Measur
esofdi
sper
sion

Ther
angegi
vesasi
mpl
eindi
cat
oroft
hev
ari
abi
l
ityofasetofobser
vat
ions.The
r
angeofasetofobser
vat
ionsi
sthedi
ff
erencebet
weent
hel
argestobser
vat
ionand
t
hesmal
l
estobser
vat
ion.

Range=hi
ghestobser
vedv
alue–l
owestobser
vedv
alue

Al
thoughi
tisv
eryeasyt
ouseandunder
stand,
ther
angei
snotar
eli
abl
ewayof

measur
ingt
hespr
eadofdat
abecausei
tisonl
ybasedononl
ytwoobser
vat
ions

whi
char
ethehi
ghestandl
owestv
alues.I
foneoft
heset
wov
aluesi
sanout
li
er,
then

t
hespr
eadofdat
aisr
atherexagger
ated.

Thev
ari
anceandst
andar
ddev
iat
ional
l
owust
oav
oidt
heshor
tcomi
ngsoft
her
ange
asameasur
eofdi
sper
sionbecauset
heyt
akei
ntoaccountal
ltheobser
vat
ionsi
n
t
hedat
asetasopposedt
ojustsel
ect
ingaf
ew.

Thev
ari
anceofasetofdat
aist
heav
eragesquar
eddev
iat
ionoft
hedat
apoi
nts

Page4of4
f
rom t
hei
rmean.Comput
ati
onal
l
y,t
hev
ari
ance( )ofasampl
eofnobser
vat
ions
i
sobt
ainedbyt
hef
ormul
a:

Thef
ormul
aforpopul
ati
onv
ari
ance i
sgi
venby
:

Thest
andar
ddev
iat
ionofasetofobser
vat
ionsi
stheposi
ti
vesquar
erootoft
he

v
ari
anceoft
heset
.

Exampl
e2

Fi
ndt
he(
a)r
angeand(
b)st
andar
ddev
iat
ionf
ort
hedat
ainExampl
e1.

Sol
uti
on2

(
a)Range=60–19=41mm

(
b)St
andar
ddev
iat
ion

Page5of5
 =121.
0488889

St
andar
ddev
iat
ion=

=11.
00mm

2.
3Hy
pot
hesi
stest
ing

Ahy
pot
hesi
sisacl
aim orasser
ti
onmadeconcer
ningapopul
ati
onpar
amet
er.A

hy
pot
hesi
stesti
sat
estusedt
over
if
ywhet
hert
hecl
aimi
sli
kel
ytobet
rueorf
alse.

I
nthi
ssect
ion,
youwi
l
lbei
ntr
oducedt
otest
sconcer
ningt
hemeanofasi
ngl
e

popul
ati
on.

Hy
pot
hesi
stest
ingi
nvol
vesgat
her
ingev
idencef
rom ar
andom sampl
edr
awnf
rom
t
hepopul
ati
onofi
nter
esti
nor
dert
odeci
dewhet
hert
henul
lhy
pot
hesi
sisl
i
kel
ytobe
t
rue orf
alse.The hy
pot
hesi
sisr
eject
ed i
fev
idence f
rom t
he sampl
eis not
consi
stentwi
tht
he st
ated hy
pot
hesi
s,ot
her
wise i
tis accept
ed.Howev
er,t
he
accept
anceoft
hest
atedhy
pot
hesi
sdoesnotnecessar
il
yimpl
ythati
tist
rue,
rat
her
i
tisar
esul
tofi
nsuf
fi
cientev
idencet
orej
ecti
t.

2.
3.1Ty
pesofhy
pot
heses
Ther
ear
etwot
ypesofhy
pot
heseswhi
char
ecal
l
edt
he:
 nul
lhy
pot
hesi
s,and
 al
ter
nat
ivehy
pot
hesi
s

Anul
lhy
pot
hesi
sisanasser
ti
onaboutt
hev
alueofapopul
ati
onpar
amet
er.I
tisa
f
ormalst
atementoft
hecl
aim bei
ngmadeconcer
ningapopul
ati
onmeasur
e.The
nul
lhy
pot
hesi
sisdenot
edbyH0.

Theal
ter
nat
ivehy
pot
hesi
s,denot
edbyH1,i
sthenegat
ionoft
henul
lhy
pot
hesi
s.For
exampl
e,anul
lhy
pot
hesi
smi
ghtasser
tthatt
hepopul
ati
onmeani
sequalt
oa
speci
fi
edv
alue .Wewr
it
ethi
sas .Theal
ter
nat
ivehy
pot
hesi
soppose

t
hisasser
ti
onandi
tiswr
it
tenas .I
nthi
scase,t
heal
ter
nat
ivehy
pot
hesi
s

Page6of6
suggest
sthatt
hemeant
akesv
aluest
hatar
eei
therbel
ow orabov
eit
.Ther
efor
e,

t
oinv
est
igat
eH0 weconductanon-
dir
ect
ionalt
estwhi
chi
sknownasat
wo-
tai
l
ed
t
est
.

Anul
lhy
pot
hesi
smi
ghtasser
tthatt
hepopul
ati
onmeani
satl
eastequalt
oacer
tai
n
speci
fi
edv
alue .Wewr
it
e .I
nthi
scase,
theal
ter
nat
ivehy
pot
hesi
swoul
d

consi
stofv
aluesbel
ow ,t
hati
s, .Si
mil
arl
y,i
fanul
lhy
pot
hesi
sasser
t

t
hatt
hepopul
ati
onmeani
slesst
hanorequalt
oaspeci
fi
edv
alue ,t
hati
s,

,t
he al
ter
nat
ive wi
l
lbe .I
n bot
hthese cases,si
nce t
he

al
ter
nat
ivehy
pot
hesesconsi
stofv
aluesei
therbel
oworabov
ethespeci
fi
edv
alue ,

weconductaone-
sidedt
estoraone-
tai
l
edt
est
.
2.
3.2Ty
peIandTy
peI
IEr
ror
s
I
ndeci
dingt
orej
ectoracceptanul
lhy
pot
hesi
s,t
her
ewi
l
lbechancesf
orer
roneousl
y
r
eject
ingoraccept
ingi
t.Sucher
ror
smaybeduet
ofaul
tysampl
i
ngpr
ocedur
es.
At
ypeIer
rori
scommi
tt
edwhenat
ruenul
lhy
pot
hesi
sisr
eject
ed.Thepr
obabi
l
ityof
commi
tt
ingat
ypeIer
rori
scal
l
edt
hel
evel
ofsi
gni
fi
canceandi
tisdenot
edby .I
tis
commont
ouse1%,
5%and10%l
evel
ofsi
gni
fi
cancei
ncal
cul
ati
ons.
At
ypeI
Ier
rori
scommi
tt
edi
fweacceptt
henul
lhy
pot
hesi
swheni
tisf
alse.The
pr
obabi
l
ityt
hatt
het
estwi
l
lbeabl
etodet
ectaf
alsenul
lhy
pot
hesi
siscal
l
edt
he
powerofat
est
.Inot
herwor
ds,t
hepowerofat
esti
sthepr
obabi
l
ityofr
eject
ingH0
wheni
ndeedH0i
sfal
se.

2.
3.3St
epsFol
l
owedi
nHy
pot
hesi
sTest
ing

Thef
oll
owi
ngst
epsshoul
dbef
oll
owedwhenconduct
ingahy
pot
hesi
stest
:

St
ep1:
Stat
etheNul
landAl
ter
nat
iveHy
pot
hesi
s
Thenul
landal
ter
nat
ivehy
pot
hesesar
especi
fi
edatt
hisi
nit
ial
stagebef
oregat
her
ing
anyev
idence.I
twoul
dbeunet
hicalandr
athermani
pul
ati
vet
ofor
mul
atet
heH0and
H1atone’
sconv
eni
enceaf
tergat
her
ingev
idence;apr
act
iset
hatwer
efert
oasdat
a
snoopi
ng.

Page7of7
St
ep2:
Ident
if
ytheDi
str
ibut
ion
Choosebet
weent
wodi
str
ibut
ionsnamel
ythez-
dist
ri
but
ionandt
het
-di
str
ibut
ion.

Whent
est
ingf
ort
hepopul
ati
onmean weuse:
a)Thez-
dist
ri
but
ionwhent
he:
 popul
ati
onst
andar
ddev
iat
ion i
sknown
 popul
ati
onst
andar
ddev
iat
ioni
sunknownand i
slar
ge(
b)Thet
-di
str
ibut
ionwhent
hepopul
ati
onst
andar
ddev
iat
ion i
sunknownand
t
hesampl
esi
ze i
ssmal
l( .
St
ep3:
Det
ermi
net
heRej
ect
ionandAccept
anceRegi
on
Dependi
ngont
hedi
str
ibut
ioni
dent
if
iedandt
hel
evel
ofsi
gni
fi
cancedesi
red,
youf
ind
av
aluef
rom st
ati
sti
calt
abl
eswhi
chwecal
lacr
it
icalv
alue.Thecr
it
icalv
alue
separ
atest
heaccept
ancer
egi
onf
rom t
her
eject
ionr
egi
on.Ther
eject
ionr
egi
oni
s
madeupofar
angeofv
aluessucht
hati
fat
estst
ati
sti
ccal
cul
atedf
rom sampl
e
dat
afal
l
sini
tthenul
lhy
pot
hesi
swoul
dber
eject
ed.Ther
eject
ionr
egi
onal
so
dependsont
henat
ureoft
heal
ter
nat
ivehy
pot
hesi
sasshowni
nthef
oll
owi
ng
f
igur
es.

ar
eaofr
eject
ion( ) ar
eaofr
eject
ion( )

cr
it
ical
val
ue 0 cr
it
ical
val
ue
Rej
ect
ionRegi
onf
or

Ar
eaofr
eject
ion

0 cr
it
ical
val
ue

Page8of8
Rej
ect
ionRegi
onf
or

ar
eaofr
eject
ion

Cr
it
ical
val
ue 0
Rej
ect
ionRegi
onf
or

St
ep4:
Cal
cul
atet
heTestSt
ati
sti
c
At
estst
ati
sti
cisav
aluecal
cul
atedf
rom sampl
edat
athati
susedt
odeci
dewhet
her
ornott
orej
ectH0.Oncet
het
estst
ati
sti
cfal
l
swi
thi
nther
eject
ionr
egi
on,H0 i
s
r
eject
ed.

Thecal
cul
ati
onoft
het
estst
ati
sti
cdependsonwhet
hert
hepopul
ati
onst
andar
d
dev
iat
ion i
sknownorunknownandal
soont
hesampl
esi
zeassummar
isedi
n
Tabl
e1bel
ow.

Tabl
e1:
TestSt
ati
sti
cforTest
ing
When i
sknown When i
sunknown
CaseI
:ni
slar
georsmal
l CaseI
I:ni
slar
ge

~N(
0,1) ~N(
0,1)

CaseI
II
:ni
ssmal
l

~t
(n-
1)

St
ep5:
Deci
deWhet
herornott
oRej
ectH0

Page9of9
Thedeci
sioni
smadeont
hebasi
sofacompar
isonbet
weent
hev
alueoft
het
est
st
ati
sti
candt
hecr
it
icalv
alue.I
fthet
estst
ati
sti
cisgr
eat
ert
hant
hecr
it
icalv
aluei
n
absol
utet
erms,
itwi
l
lfal
li
nther
eject
ionr
egi
ont
husl
eadi
ngt
other
eject
ionofH0.

St
ep6:
MakeaConcl
usi
on
I
fH0 i
srej
ect
ed,weconcl
udet
hatH1 i
spr
obabl
ytr
ue.I
fwef
ailt
orej
ectH0,we
concl
udet
hatt
heev
idencegat
her
edi
sinsuf
fi
cientt
owar
rantt
her
eject
ionofH0.

Exampl
e3

Ani
mmi
grat
ionof
fi
cercl
aimst
hatatl
east40peopl
ecr
osst
hebor
deri
l
legal
l
yint
o
Sout
hAf
ri
caev
ery
day
.Tot
estt
hecl
aim,
thenumberofpeopl
ecr
ossi
ngi
l
legal
l
yint
o
Sout
hAf
ri
cawasnot
edon12r
andoml
ysel
ect
edday
sasf
oll
ows:
284136501739216426304212

Testt
hecl
aim madebyt
hei
mmi
grat
ionof
fi
cerata5%l
evel
ofsi
gni
fi
cance.

Sol
uti
on3
1.

2.Thepopul
ati
onst
andar
ddev
iat
ion i
sunknown,
butt
hesampl
esi
ze =12i
s
smal
l
,soweuset
het
-di
str
ibut
ion.

3. and i
t i
s a one-
tai
l
ed t
est
. The cr
it
ical v
alue i
s

-
1.80 0
Wewoul
drej
ectH0i
f 1.
80

Page10of10
 4.Youshoul
dver
if
ythat ands=14.
6339

=-
0.2762

5.Si
nce =0.
2762<1.
80,
wef
ail
tor
ejectH0

6.Thedat
adoesnotpr
ovi
desuf
fi
cientev
idenceat5% l
evelofsi
gni
fi
cancet
o
r
ejectH0,
ther
efor
eweconcl
udet
hatt
hecl
aimi
sli
kel
ytobet
rue.

3.
0Bi
var
iat
eSt
ati
sti
cs
Bi
var
iat
eSt
ati
sti
csr
efer
stost
ati
sti
csper
tai
ningt
otwov
ari
abl
es.
3.
1Cor
rel
ati
onanal
ysi
s
Cor
rel
ati
onanal
ysi
sisast
ati
sti
calpr
ocedur
ethati
susedt
omeasur
etheext
entt
o
whi
cht
wov
ari
abl
esar
erel
ated.Twov
ari
abl
esar
ehi
ghl
ycor
rel
atedi
ftheymov
ewel
l
t
oget
hersucht
hati
tispossi
blet
opr
edi
ctt
hemov
ementofonev
ari
abl
eusi
ng
knowl
edgeaboutt
hemov
ementoft
heot
herv
ari
abl
e.Forexampl
e,goodr
ainsar
e
associ
atedwi
thgoodhar
vest
s(ot
hercondi
ti
onshel
dconst
ant
).
Thecor
rel
ati
oncoef
fi
cientdenot
edby i
sanumer
icalmeasur
eoft
hest
rengt
hof
t
hel
i
nearr
elat
ionshi
pbet
weent
wov
ari
abl
esXandY.Acor
rel
ati
oncoef
fi
cientt
akes
v
aluesbet
ween-
1and+1i
ncl
usi
ve,
thati
s .

Page11of11
Possi
blev
aluesof ar
eint
erpr
etedasf
oll
ows:
 i
sequal
tozer
o,i
ndi
cat
est
her
eisnocor
rel
ati
onbet
weent
hev
ari
abl
es.
 i
ndi
cat
esaper
fectposi
ti
vecor
rel
ati
onbet
weent
hev
ari
abl
es.
 i
ndi
cat
esaper
fectnegat
ivecor
rel
ati
onbet
weent
hev
ari
abl
es
 cl
oset
o+1i
ndi
cat
esast
rongposi
ti
vecor
rel
ati
on
 cl
oset
o-1i
ndi
cat
esast
rongnegat
ivecor
rel
ati
on
 cl
oset
ozer
oimpl
i
esaweakcor
rel
ati
onbet
weent
hev
ari
abl
es

Thef
oll
owi
ngadj
ect
ivesmayhel
pyout
odescr
ibet
hedegr
eeofl
i
nearassoci
ati
on
bet
weent
wov
ari
abl
es:

Val
uesof Sui
tabl
eadj
ect
ives
+0.
7to+1.
0 St
rong,
posi
ti
ve
+0.
4to+0.
69 Fai
r/
moder
ate,
posi
ti
ve
+0.
3to+0.
39 Weak,
posi
ti
ve
0.
0to+0.
29 Negl
i
gibl
e/scantposi
ti
ve
0.
0to-
0.29 Negl
i
gibl
e/scantnegat
ive
-
0.3t
o-0.
39 Weak,
negat
ive
-
0.4t
o-0.
69 Fai
r/
moder
ate,
negat
ive
-
0.7t
o-1.
0 St
rong,
negat
ive

Ther
ear
etwocommonl
yusedcor
rel
ati
oncoef
fi
cient
swhi
char
e:
a)Pear
son’
spr
oductmomentcor
rel
ati
oncoef
fi
cient
()
b)Spear
man’
srankcor
rel
ati
oncoef
fi
cient( )

3.
1.1Pear
son’
spr
oductmomentcor
rel
ati
oncoef
fi
cient
Pear
son’
scor
rel
ati
oncoef
fi
cienti
scal
cul
atedf
orr
ati
odat
aori
nter
valdat
a.I
tis
basedont
hemeanandst
andar
ddev
iat
ionandt
her
efor
ecanbeaf
fect
edbyext
reme
v
alues.Pear
son’
scoef
fi
cienti
sobt
ainedusi
ngt
hef
oll
owi
ngcomput
ati
onal
for
mul
a:

Page12of12
Exampl
e4
Thef
oll
owi
ngdat
aar
ear
andom sampl
eofi
ndexedpr
icesofgol
dandpl
ati
num ov
er
asi
xyearper
iod:

Gol
d(X) 12 10 14 11 12 9
Pl
ati
num (
Y) 18 17 23 19 20 15

Cal
cul
ateandi
nter
prett
hecor
rel
ati
oncoef
fi
cientf
ort
hedat
a.

Sol
uti
on4

Acor
rel
ati
oncoef
fi
ci
entof0.
9474i
scl
oset
o+1andi
tindi
cat
esast
rongposi
ti
ve
cor
rel
ati
onbet
weent
hepr
icesofgol
dandpl
ati
num.Whenpl
ati
num pr
icesar
egoi
ng
up,
gol
dpr
icesar
eal
soexpect
edt
ogoup,
othert
hingsbei
ngequal
.

3.
1.2Spear
man’
srankcor
rel
ati
oncoef
fi
cient
TheSpear
man’
scor
rel
ati
oncoef
fi
cientdenot
edby i
scal
cul
atedf
orr
ankeddat
a.

Thi
sis dat
a measur
ed on t
he or
dinalscal
e.The cor
rel
ati
on coef
fi
cient i
s

i
nter
pret
ed i
nthe same way as t
he Pear
son’
s cor
rel
ati
on coef
fi
cient .The
comput
ati
onal
for
mul
afor i
sgi
venby

Page13of13
wher
e i
sthedi
ff
erencei
nranksobt
ainedbysubt
ract
ingt
her
anksofyv
aluesf
rom
t
her
anksofxv
aluesf
oreachpai
rofobser
vat
ions.

Exampl
e5
Apanel
oft
woj
udgesr
ankedt
hequal
i
tyofpot
atoeshar
vest
edf
rom 5A1f
arms(
A,B,
C,
DandE)asf
oll
ows:

A1f
arm A B C D E
Judge1 4 5 1 3 2
Judge2 5 4 2 3 1

I
sther
eagr
eementi
nthemanneri
n whi
ch t
hej
udgesper
cei
vet
hequal
i
tyof
pot
atoesf
rom t
hef
ivef
arms?
Sol
uti
on5
Judge1 Judge2 di di2
4 5 -
1 1
5 4 1 1
1 2 -
1 1
3 3 0 0
2 1 1 1

=0.
8

Thecor
rel
ati
oncoef
fi
ci
enti
sfai
rl
yhi
ghandposi
ti
veshowi
ngt
hatt
hej
udgesdonot

Page14of14
di
ff
ermuchi
nthewayt
heyper
cei
vet
hequal
i
tyofpot
atoesf
rom t
he5f
arms.

3.
2Regr
essi
onAnal
ysi
s
I
nreall
i
fesi
tuat
ionswear
eusual
l
yint
erest
edi
nrel
ati
onshi
psbet
weenv
ari
abl
es.
Si
mpl
eli
nearr
egr
essi
oni
susedt
oest
abl
i
sht
hef
unct
ionalf
orm ofar
elat
ionshi
p
bet
weent
wov
ari
abl
es.

3.
2.1Ty
pesofVar
iabl
es
Ther
ear
etwo t
ypesofv
ari
abl
est
hatwedealwi
thi
nsi
mpl
eli
nearr
egr
essi
on
anal
ysi
s.Thev
ari
abl
ethathasadependencer
elat
ionshi
pont
heot
herv
ari
abl
eis
cal
l
edt
her
esponsev
ari
abl
eordependentv
ari
abl
e,denot
edbyY.Theot
herv
ari
abl
e
i
scal
l
edt
heexpl
anat
oryori
ndependentv
ari
abl
edenot
edbyX.

3.
2.2Scat
terPl
ots

Random pai
rsofobser
vat
ionsf
ort
hei
ndependentanddependentv
ari
abl
esar
e
pl
ott
edonagr
apht
oshowt
heki
ndofr
elat
ionshi
pthatexi
stsbet
weent
hev
ari
abl
es.
Fr
om t
hescat
terpl
oty
oucant
elli
fther
elat
ionshi
pcanbemodel
l
edbyast
rai
ght
l
i
neequat
ionbef
orepr
oceedi
ngt
odol
i
nearr
egr
essi
onanal
ysi
s.

Exampl
e6
Ar
esear
cherwoul
dli
ket
oest
abl
i
sht
hei
mpactofpopul
ati
ongr
owt
honeconomi
c
dev
elopment
.He col
l
ect
ed t
he f
oll
owi
ng cr
oss-
sect
ionaldat
afr
om si
x SADC
count
ri
es.
Popul
ati
ongr
owt
hrat
e(%) Annual
GDP(
US$bi
l
li
on)
7 18
Dr
aw a 9 35 scat
ter pl
ot
t
o 5 12 r
epr
esentt
he
dat
a. 15 50 Commenton
t
he ki
nd 12 36 of

6 24 r
elat
ionshi
p
bet
ween popul
ati
on
gr
owt
handeconomi
cdev
elopment
.
Page15of15
Sol
uti
on6

Thepoi
ntsseem t
obef
oll
owi
ngal
i
newi
thposi
ti
vegr
adi
ent
.Ify
oui
nser
tal
i
neof
bestf
itt
hrought
hepoi
nts,
youwi
l
lseet
hatt
hepoi
ntsdonotdev
iat
emuchf
rom t
he
l
i
ne.Wecant
her
efor
econcl
udet
hatt
her
eisast
rongposi
ti
vel
i
nearr
elat
ionshi
p
bet
weenpopul
ati
ongr
owt
handeconomi
cdev
elopment
.

3.
2.3TheSi
mpl
eLi
nearRegr
essi
onModel

Thesi
mpl
eli
nearr
egr
essi
onmodel
isgi
venby

wher
e i
sthedependentv
ari
abl
e, i
sthei
ndependentv
ari
abl
e, i
stheer
rort
erm,
and ar
ecal
l
edpopul
ati
onr
egr
essi
oncoef
fi
cient
srepr
esent
ingt
hei
nter
cept

and sl
oper
espect
ivel
y.Themodelr
epr
esent
stheunknownpopul
ati
onr
elat
ionshi
p
bet
weent
het
wov
ari
abl
es.Theer
rort
ermi
sincl
udedi
nthemodelt
ocapt
uret
he
ef
fectofot
heri
mpor
tantv
ari
abl
est
hatar
eomi
tt
edi
nthemodel
.

Page16of16
Ther
egr
essi
onequat
ioni
sest
imat
edby
:

wher
e

and

i
stheest
imat
edr
egr
essi
onequat
ionconnect
ingv
ari
abl
esXandY,
wher
e
and ar
eest
imat
esoft
hepopul
ati
oni
nter
cept andpopul
ati
onsl
ope oft
he

l
i
ner
espect
ivel
y.

Exampl
e7
Est
imat
ether
egr
essi
onequat
ionf
ort
hedat
aofExampl
e6

Sol
uti
on7
Usi
ng t
het
wo v
ari
abl
est
ati
sti
calmodeon y
ourcal
cul
ator
,you wi
l
lobt
aint
he
f
oll
owi
ngr
esul
ts:

=3.
405405405

=-
1.481981978

Ther
egr
essi
onequat
ioni
s

Page17of17
3.
2.4I
nter
pret
ati
onof and
Thepar
amet
er i
sthei
nter
ceptont
hedependentv
ari
abl
e.I
tist
hev
aluet
hatt
he
v
ari
abl
e i
spr
edi
ctedt
oassumei
fthev
ari
abl
e hasazer
oval
ue.Youshoul
d
guar
dagai
nsti
nter
pret
ingt
hei
nter
cepti
nter
msoft
hedependentv
ari
abl
eift
he
r
angeof -
val
uesusedt
oconst
ructt
hemodel
donoti
ncl
udezer
o.

Thepar
amet
er r
epr
esent
sther
ateofchangeof wi
thr
espectt
o .Thust
he
v
alueof showst
hecor
respondi
ngchangei
nthev
alueof f
orev
eryuni
tchangei
n
t
hev
alueof .

I
ntheequat
ionobt
ainedabov
e,t
hev
alue cannotbei
nter
pret
edi
n
t
ermsofeconomi
cdev
elopmentsi
ncet
he v
aluesusedt
oconst
ructt
heequat
ion
di
dnoti
ncl
udezer
o.
Thev
alue r
epr
esent
sthecor
respondi
ngi
ncr
easei
nGDPf
orev
ery
uni
tincr
easei
nthepopul
ati
ongr
owt
hrat
e.

3.
2.5Est
imat
ingVal
uesoft
heDependentVar
iabl
e
Ther
egr
essi
onmodelcanbeusedt
oest
imat
eval
uesoft
hedependentv
ari
abl
e
gi
venv
aluesoft
hei
ndependentv
ari
abl
e.Agi
venv
alueofXi
ssubst
it
utedi
ntot
he
r
egr
essi
onequat
iont
oobt
aint
hecor
respondi
ngv
alueofY.Themodelcangi
ve
r
eli
abl
eest
imat
esofv
aluesofYf
orX-
val
ueswi
thi
nther
angeofv
aluesusedt
o
const
ructt
hemodel
.Out
sidet
her
angeofX-
val
uesusedi
ntheconst
ruct
ionoft
he
model
,themodelbecomesunpr
edi
ctabl
eandmaygi
vemi
sleadi
ngr
esul
ts.Fort
his
r
eason,
youshoul
dguar
dagai
nstext
rapol
ati
oni
nregr
essi
onanal
ysi
s.

Exampl
e8
Uset
hemodel
obt
ainedi
nExampl
e7t
oest
imat
ethel
evel
ofeconomi
cdev
elopment
i
fthepopul
ati
onofacount
rygr
owsby10%.

Sol
uti
on8

=US$32.
57207207bi
l
li
on

Page18of18
Page19of19