Measurement: scaling, reliability and validity

8/18/2019 Measurement: scaling, reliability and validity

1/42

MEASUREMENT: SCAL

RELIABILITY, VALI

GROUP

AHMED JAN DAHRI (

ASGHAR ALI ARAIN (


2/42

MEASUREMENT: SCALING, RELIABILITY, VALIDI

• 1. Rating Scales

Dicht!"s scale

Categ#$ scale

Li%e#t scale

N"!e#ical scales

Se!antic &i''e#ential scaleIte!i(e& #ating scale

)i*e& # cnstant s"! #ating scale

Sta+le scale

G#a+hic #ating scale

Cnsens"s scale


3/42

• 1.1 DIC-T-M-US SCALE

The &icht!"s scale is "se& t elicit a Yes # N anse#, as in e*a

/el. Nte that a n!inal scale is "se& t elicit the #es+nse.

E*a!+le 0.1D $" n a ca# Yes N


4/42

• 1.2 Categ#$ Scale

The categ#$ scale "ses !"lti+le ite!s t elicit a single #es+nse as +e# th

'lling e*a!+le. This als "ses n!inal scale.

E*a!+le 0.2

3he#e in n#the#n Cali'#nia & $" #esi&e 4444N#th Ba$

4444S"th Ba$

4444East Ba$

44445enins"la

4444-the#


5/42

• 1.6. Li%e#t Scale

The Li%e#t scale is &esigne& t e*a!ine h st#ngl$ s"/7ects ag#ee

# &isag#ee ith state!ents n a 89+int scale ith the 'lling an


6/42

• The #es+nses e# a n"!/e# ' ite!s ta++ing a +a#tic"la# cnce+t

a#ia/le ;as +e# the 'lling e*a!+le< a#e then s"!!ate& '# ee#

#es+n&ent. This is an inte#al scale an& the &i''e#ences in the #es+

/eteen an$ t +ints n the scale #e!ain the sa!e.


7/42

• 1.= Se!antic Di''e#ential Scale

See#al /i+la# att#i/"tes at the i&enti'ie& at the e*t#e!es ' the scale, a

#es+n&ents a#e as%e& t in&icate thei# attit"&es, n hat !a$ /e calle&

se!antic s+ace, ta#& a +a#tic"la# in&ii&"al, /7ect, # eent n each

att#i/"tes. The /i+la# a&7ecties "se&, '# instance, "l& e!+l$ s"ch

as G&999Ba&> St#ng?ea%> t?Cl&. The #es+nses can /e +ltte/tain a g& i&ea ' thei# +e#ce+tins. This is t#eate& as an inte#al sca


8/42

• E*a!+le 0.=


9/42

• 1.8. N"!e#ical Scale

The n"!e#ical scale is si!ila# t the se!antic &i''e#ential scale, ith the

&i''e#ence that n"!/e#s n a 89+int # @9+int scale a#e +#i&e&, ith /

a&7ecties at /th en&s, as ill"st#ate& /el. This is als an inte#al scale.

E*a!+le 0.8


10/42

• 1. Ite!i(e& Rating Scale

A 89+int # @9+int scale ith anch#s, as nee&e&, is +#i&e& '# each ite!

the #es+n&ent states the a++#+#iate n"!/e# n the si&e ' each ite!, # ci

the #eleant n"!/e# against each ite!. The #es+nses t the ite!s a#e then

s"!!ate&. This "ses an inte#al scale.

• E*a!+le 0. ;I<

Res+n& t each ite! "sing the scale /el, an& in&icate $"# #es+nse n"

n the line /$ each ite!.


11/42

• E*a!+le 0. ;ii<

Ci#cle the n"!/e# that is clsest t h $" 'eel '# ite! /el.


12/42

• 1.@ )i*e& # Cnstant S"! Scale

The #es+n&ents a#e he#e as%e& t &ist#i/"te a gien n"!/e# ' +ints

ac#ss a#i"s ite!s as +e# the e*a!+le /el. This is !#e in the nat"#e

an #&inal scale.

E*a!+le 0.@

In chsing a tilet sa+, in&icate the i!+#tance $" attach t each ' th'lling 'ie as+ects /$ alltting +ints '# each t ttal 1 in all.


13/42

• 1. Sta+el Scale

This scale si!"ltane"sl$ !eas"#es /th the &i#ectin an& intensit$

' the attit"&e ta#& the ite!s "n&e# st"&$. The cha#acte#istic ' inte#es

t the st"&$ is +lace& at the c#ne# an& a n"!e#ical scale #anging, sa$, '#

6 t 96, n eithe# si&e ' the ite! as ill"st#ate& /el. This gies an i&ea

h clse # &istant the in&ii&"al #es+nse t the sti!"l"s is, as shn in

the e*a!+le /el. Since this &es nt hae an a/sl"te (e# +int, this isan inte#al scale.


14/42

• E*a!+le 0.


15/42

• 1.0 G#a+hic Rating Scale

A g#a+hical #e+#esentatin hel+s the #es+n&ents t in&icate n this

scale thei# anse#s t a +a#tic"la# "estin /$ +lacing a !a#% at the

a++#+#iate +int n the line, as in the 'lling e*a!+le. This is an

#&inal scale, th"gh it !ight a++ea# t !a%e it l% li%e an inte#al

scale.


16/42

• E*a!+le 0.0


17/42

• 1.1. Cnsens"s Scale

Scales a#e als &eel+e& /$ cnsens"s, he#e a +anel ' 7"&ges selects

ce#tain ite!s, hich in its ie !eas"#e the #eleant cnce+t. The ite!s a#e

chsen +a#tic"la#l$ /ase& n thei# +e#tinence # #eleance t the cnce+t. -ne

s"ch cnsens"s scale is the Th"#stne E"al A++ea#ing Inte#al Scale,

he#e a cnce+t is !eas"#e& /$ a c!+le* +#cess 'lle& /$ a +anel '

7"&ges. Using a +ile ' ca#&s cntaining see#al &esc#i+tins ' the cnce+t, a

+anel ' 7"&ges ''e#s in+"ts t in&icate h clse # nt the state!ents a#e t

the cnce+t "n&e# st"&$. The scale is then &eel+e& /ase& n the cnsens"s

#eache&.


18/42

• 1.11. -the# Scales

The#e a#e als s!e a&ance& scaling !eth&s s"ch as !"lti&i!ensinal

scaling, he#e /7ects, +e+le, # /th, a#e is"all$ scale&, an& a cn7int

anal$sis is +e#'#!e&. This +#i&es a is"al i!age ' the #elatinshi+s in

s+ace a!ng the &i!ensins ' a cnst#"ct.


19/42

• 2. Ran%ing Scales

Ran%ing scales a#e "se& t ta+ +#e'e#ences /eteen t # a!ng !#e

/7ects # ite!s ;#&inal in nat"#e


20/42

• 2.2 )#ce& Chice

The '#ce& chice ena/les #es+n&ents t #an% /7ects #elatie t ne

anthe#, a!ng the alte#naties +#i&e&. This is easie# '# the #es+n&ents,

+a#tic"la#l$ i' the n"!/e# ' chices t /e #an%e& is li!ite& in n"!/e#.

• E*a!+le 0.1

Ran% the 'lling !aga(ines that $" "l& li%e t s"/sc#i/e t in the#&e# ' +#e'e#ence, assign!ent 1 '# the !st +#e'e##e& chice an& 8 '# the lea

+#e'e##e&.


21/42

• 2.6 C!+a#atie Scale

The c!+a#atie scale +#i&es a /ench!a#% # a +int ' #e'e#ence t

assess attit"&es ta#&s the c"##ent /7ect, eent, # sit"atin "n&e# st"&$.

An e*a!+le ' the "se ' c!+a#atie scale 'lls.

E*a!+le 0.11

Rating scales a#e "se& t !eas"#e !st /ehai#al cnce+ts. Ran%ing scales

a#e "se& t !a%e c!+a#isns # #an% the a#ia/les that hae /een ta++e& n

a n!inal scale.


22/42

• 6. G&ness ' Meas"#es

It is i!+#tant t !a%e s"#e that the inst#"!ent that the inst#"!ent that e

&eel+ t !eas"#e a +a#tic"la# cnce+t is inst#"!ent that e &eel+ t

!eas"#es a +a#tic"la# cnce+t is in&ee& acc"#atel$ !eas"#ing the a#ia/le,

an& that in 'act, e a#e act"all$ !eas"#ing the cnce+t that e set "t t

!eas"#e. This ens"#es that in +e#atinall$ &e'ining +e#ce+t"al an&

attit"&inal a#ia/les, e hae nt e#l%e& s!e i!+#tant &i!ensins an&

ele!ents # incl"&e& s!e i##eleant nes.


23/42

• 6.1. Ite! Anal$sis

Ite! anal$sis is &ne t see i' the ite!s in the inst#"!ent /elng the#e # nt.

Each ite! is e*a!ine& '# its a/ilit$ t &isc#i!inate /eteen thse s"/7ects

hse ttal sc#es a#e high, an& thse ill l sc#es. In ite! anal$sis, the

!eans /eteen the high9sc#e g#"+ an& the l9sc#e g#"+ a#e teste& t

&etect signi'icant &i''e#ences th#"gh the t9al"es. The ite!s ith a high t9al"hich is a/le t i&enti'$ the highl$ &isc#i!inating ite!s in the

inst#"!ent< a#e then incl"&e& in the inst#"!ent.


24/42

• =. Relia/ilit$

The #elia/ilit$ ' a !eas"#e in&icates the e*tent t hich it is ith"t /ias

;e### '#ee< an& hence ens"#es cnsistent !eas"#e!ent ac#ss ti!e an& ac#ss

the a#i"s ite!s in the inst#"!ent. In the# #&s, the #elia/ilit$ ' a

!eas"#e is an in&icatin ' the sta/ilit$ an& cnsistenc$ ith hich the

inst#"!ent !eas"#es the cnce+t an& hel+s t assess the Fg&ness ' a!eas"#e.


25/42

• =.1 Sta/ilit$ ' Meas"#es

The a/ilit$ ' a !eas"#e t #e!ain the sa!e e# ti!e?&es+ite "ncnt#lla/le

testing cn&itins # the state ' the #es+n&ents the!seles?is in&icatie '

its sta/ilit$ an& l "lne#a/ilit$ t changes in the sit"atin. This attests t its

Fg&ness /eca"se the cnce+t is sta/l$ !eas"#e&, n !atte# hen it is &ne.

T tests ' sta/ilit$ a#e test9#etest #elia/ilit$ an& +a#allel9'#! #elia/ilit$.


26/42


27/42

• =.2 Test9Retest Relia/ilit$

The #elia/ilit$ ce''icient /taine& ith a #e+etitin ' the sa!e !eas"#e n

a secn& ccasin is calle& test9#etest #elia/ilit$. That is, hen a

"estinnai#e is a&!iniste#e& t a set ' #es+n&ents n, an& again t the

sa!e #es+n&ents, sa$s see#al ee%s t !nths late#, then the c##elatin

/eteen the sc#es /taine& at the t &i''e#ent ti!es '#! ne an& the

sa!e set ' #es+n&ents is calle& the test9#etest ce''icient. The highe# it is,

the /ette# the test9#etest #elia/ilit$, an& cnse"entl$, the sta/ilit$ ' the

!eas"#e ac#ss ti!e.


28/42

• =.6 5a#allel9)#! Relia/ilit$

3hen #es+nses n t c!+a#a/le sets ' !eas"#es ta++ing the sa!e

cnst#"ct a#e highl$ c##elate&, e hae +a#allel9'#! #elia/ilit$. Bth '#!s

hae si!ila# ite!s an& the sa!e #es+nse '#!at, the nl$ changes /eing the

#&ings an& the #&e# # se"ence ' the "estins. 3hat e t#$ t

esta/lish he#e is the e### a#ia/ilit$ #es"lting '#! #&ing an& #&e#ing ' the "estins. I' t s"ch c!+a#a/le '#!s a#e highl$ c##elate& the

!eas"#es a#e #easna/l$ #elia/le.


29/42

• =.= Inte#nal Cnsistenc$ ' Meas"#es

The inte#nal cnsistenc$ ' !eas"#es is in&icatie ' the h!geneit$ ' the

ite!s in the !eas"#e that ta+ the cnst#"ct. In the# #&s, the ite!s sh"l&

Fhang tgethe# as a set, an& /e ca+a/le ' in&e+en&entl$ !eas"#ing the

sa!e cnce+t s that the #es+n&ents attach the sa!e e#all !eaning t

each ' the ite!s. This can /e seen /$ e*a!ining i' the ite!s an& the s"/sets

' ite!s in the !eas"#ing inst#"!ents a#e c##elate& highl$. Cnsistenc$ can/e e*a!ine& th#"gh the inte#9ite! cnsistenc$ #elia/ilit$ an& s+lit9hal'

#elia/ilit$ tests.


30/42

• a. Inte#ite! Cnsistenc$ Relia/ilit$

This is a test ' the cnsistenc$ ' #es+n&entsH anse#s t all the ite!s in a

!eas"#e. T the &eg#ee that ite!s a#e in&e+en&ent !eas"#es ' the sa!e

cnce+t, the$ ill /e c##elate& ith ne anthe#. The !st ++"la# test '

inte#ite! cnsistenc$ #elia/ilit$ is the C#n/achHs ce''icient al+ha

;C#n/achHs al+ha> C#n/ach, 10=


31/42

• /. S+lit9al' Relia/ilit$

S+lit9hal' #elia/ilit$ #e'lects the c##elatins /eteen t hales ' an

inst#"!ent. The esti!ates "l& a#$ &e+en&ing n h the ite!s in the

!eas"#e a#e s+lit int t hales. S+lit9hal' #elia/ilities c"l& /e highe# than

C#n/achHs al+ha nl$ in the ci#c"!stance ' the#e /eing !#e than ne

"n&e#l$ing #es+nse &i!ensin ta++e& /$ the !eas"#e an& hen ce#tainthe# cn&itins a#e !et as ell. ence, in al!st all cases, C#n/achHs al+ha

can /e cnsi&e#e& a +e#'ectl$ a&e"ate in&e* ' the inte#ite! cnsistenc$

#elia/ilit$.


32/42

• 8. Vali&it$

See#al t$+es ' ali&it$ tests a#e "se& t test the g&ness ' !eas"#es an&

#ite#s "se &i''e#ent te#!s t &ente the!. )# the sa%e ' cla#it$, e !a$

g#"+ ali&it$ tests "n&e# th#ee /#a& hea&ings: cntent ali&it$,c#ite#in9#elate& ali&it$, an& cnst#"ct ali&it$.

8.1 Cntent Vali&it$

Cntent ali&it$ ens"#es that the !eas"#e incl"&es an a&e"ate an&

#e+#esentatie set ' ite!s that ta+ the cnce+t. The !#e the scale ite!s#e+#esent the &!ain # "nie#se ' the cnce+t /eing !eas"#e&, the g#eate

the cntent ali&it$. T +"t it &i''e#entl$, cntent ali&it$ is a '"nctin ' h

ell the &i!ensins an& ele!ents ' a cnce+t hae /een &elineate&.

)ace ali&it$ is cnsi&e#e& /$ s!e as a /asic an& a e#$ !ini!"! in&e*

cntent ali&it$. )ace ali&it$ in&icates that the ite!s that a#e inten&e& t

!eas"#e a cnce+t, & n the 'ace ' it l% li%e the$ !eas"#e the cnce+t.


33/42

• 8.2 C#ite#in9Relate& Vali&it$

C#ite#in9#elate& ali&it$ is esta/lishe& hen the !eas"#e &i''e#entiates

in&ii&"als n a c#ite#in it is e*+ecte& t +#e&ict. This can /e &ne /$

esta/lishing cn9c"##ent ali&it$ # +#e&ictie ali&it$, as e*+laine& /el

Cnc"##ent ali&it$ is esta/lishe& hen the scale &isc#i!inates in&ii&"a

a#e %nn t /e &i''e#ent> that is, the$ sh"l& sc#e &i''e#entl$ n theinst#"!ent as in the e*a!+le that 'lls.


34/42

• 8.6 Cnst#"ct Vali&it$

Cnst#"ct ali&it$ testi'ies t h ell the #es"lts /taine& '#! the "se

!eas"#e 'it the the#ies a#"n& hich the test is &esigne&. This is assesse&

th#"gh cne#gent an& &isc#i!inant ali&it$, hich a#e e*+laine& /el

Cne#gent ali&it$ is esta/lishe& hen the sc#es /taine& ith t &i''

inst#"!ents !eas"#ing the sa!e cnce+t a#e highl$ c##elate&.Disc#i!inant Vali&it$ is esta/lishe& hen, /ase& n the#$, t a#ia/les

+#e&icte& t /e "nc##elate&, an& the sc#es /taine& /$ !eas"#ing the!

in&ee& e!+i#icall$ '"n& t /e s.


35/42

MEASURES FROM MANAGEMENT


36/42

MEASURES FROM MANAGEMENTRESEARCH• Bel! "# $ #$%&le ' e #*$le# +#e, %e$#+.e e $."$/le# .el$e,

%$$e%e .e#e$.*2


37/42


38/42


39/42


40/42

4 LEAST PREFERRED COOR6ER SCALE (TO ASHETHER EMPLO7EES ARE PRIMARIL7 PEOPLE-ORIENTED OR TAS6-ORIENTED)

• L% at the #&s at /th en&s ' the line /e'#e $" +"t in $"# KK. 5lease #e!‖

the#e a#e n #ight # #ng anse#s. 3#% #a+i&l$> $"# 'i#st anse# is li%el$ t /

5lease & nt !it an$ ite!s, an& !a#% each ite! nl$ nce.

L5C

Thin% ' the +e#sn ith h! $" can #% least ell. e !a$ /e s!ene $"

n, # he !a$ /e s!ene $" %ne in the +ast.

e &es nt hae t /e the +e#sn $" li%e least ell, /"t sh"l& /e the +e#sn ith

$" ha& the !st &i''ic"lt$ in getting a 7/ &ne. Desc#i/e this +e#sn as he a++ea


41/42


42/42

AN7 8UESTION IF

Measurement: scaling, reliability and validity

Documents