Newsom 1

USP 656 Multilevel Regression

Winter 2006
jhlhkhhhkjhkjhkjkjhkhkhkjhk (i.e., 0 mg, 5 mg, n! "0 mg#. $owever, i% the reser&her wnts to mke
in%eren&es 'e(on! the )rti&ulr vlues o% the %g%%%%%%%%%%%%%%%%%%%%%%%%%%kkkkk!%gh!gh'le use! in the stu!(,
rn!om e%%e&ts mo!el is use!. * &ommon e+m)le woul! 'e the use o% )u'li& rt works re)resenting
low, mo!erte, n! &v'&'&'&v'&v'
Fixed and Random Coefficients in Multilevel Regression
,he rn!om vs. %i+e! !hjkghjkjkjkghkjkjkjkjkn! e%%e&ts is im)ortnt in multilevel regression. -n
multilevel regression mo!els, 'oth level." &v&v'&v'v&' to vr( rn!oml( &ross grou)s. /e&use o%
the ssum)tions 'out their error !istri'utions, we &ll their vrin&es, oo=var(U0j) n! 11=var(U1j),
0rn!om &oe%%i&ients.1
,his mens tht we &n think 'out 0j n! 1j s kin to the rn!om variables - !es&ri'e!
'ove. -nste! o% ttem)ting to generli2e 'e(on! the )rti&ulr vlues o% the in!e)en!ent vri'le,
we re ttem)ting to generli2e 'e(on! the )rti&ulr grou)s in the stu!(. 3or instn&e, we m( hve
50 &om)nies, 'ut we wish to generli2e to lrger universe o% &om)nies when we e+mine the
mens (inter&e)ts# or the +.( reltionshi) (slo)es#.
-n $4M, vrin&es %or the inter&e)ts n! slo)es re estimte! '( !e%ult. ,ht is, when (ou
%irst &onstru&t mo!el, U0j n! U1j re estimte! '( !e%ult. $owever, the reser&her hs the &hoi&e
o% setting U0j n! U1j to 'e 2ero. /( setting them to 2ero, we re testing mo!el in whi&h we ssume
0j n! 1j !o not vr( rn!oml( &ross grou)s. -n %&t, their vrin&e is ssume! to 'e 2ero, so the(
re ssume! to 'e &onstnt or 0nonvr(ing1 &ross grou)s. 3or e+m)le, %i+e!, nonvr(ing inter&e)ts
woul! im)l( the grou) verge %or the !e)en!ent vri'le is ssume! to 'e e5ul in e&h grou).
*lthough we t()i&ll( re%er to this &onstrint s 0%i+ing the inter&e)ts1 or 0%i+ing the slo)es,1 the term is
somewht loosel( ))lie!, 'e&use we re rell( ssuming the( re %i+e! and nonvarying.
Intercept Model and Random Effects ANOVA
-n the sim)lest $4M mo!el, there re no )re!i&tors. We hve one level." e5ution n! one level.2
e5ution6
0 ij j ij
Y R = +
where Y is the !e)en!ent vri'le, 0j is the inter&e)t, n! Rij is the resi!ul or error. 0j &n 'e
thought o% s the men o% e&h grou). ,he level.2 e5ution lso hs no )re!i&tors in it7s sim)lest
%orm6
0 00 0 j j
U = +
where
0j
is the !e)en!ent vri'le,
00
is the level.2 inter&e)t, n! U
0j
is the level.2 error. -n this
e5ution, 00 re)resents the grn! men or the men o% the inter&e)ts. U0j re)resents the !evition o%
e&h men %rom the grn! men. When the verge !evition is lrge, there re lrge grou)
!i%%eren&es.
-t &n 'e shown tht this is rell( the *N89* mo!el. *N89* e+mines the !evitions o% grou) mens
%rom the grn! men. /e&use we ssume tht the grou) mens, re)resente! '( 0j n!, thus, their
!evitions re rn!oml( vr(ing, this mo!el is e5uivlent to the random effects *N89* mo!el. We
&n rewrite our two $4M e5utions s single e5ution, i% we )lug in the level.2 terms into level."
e5ution.
00 0 ij j ij
Y U R = + +
-n this e5ution, 00 re)resents the grn! men, the U0j term re)resents the !evitions %rom the men
(or men !i%%eren&es or tretment e%%e&t, i% (ou woul! like#, n! Rij re)resents the within.grou) error or
vrition. -n more mthemti&ll( oriente! *N89* te+t'ooks, (ou will see n *N89* mo!el written
something like6
Newsom 2
USP 656 Multilevel Regression
Winter 2006
ij j ij
Y e = + +
-% the mo!el is n rn!om e%%e&ts *N89*, j is ssume! to hve rn!om !istri'ution. -n multilevel
regression, 'e&use we ssume U0j to vr( rn!oml(, the sim)le $4M mo!el with no level." or level.2
)re!i&tors is e5uivlent to the rn!om e%%e&ts *N89* mo!el.
When we !! )re!i&tors to the level." e5ution, the( re &ovrites n! the mo!el 'e&omes
rn!om e%%e&ts *N:89* in whi&h the mens re !juste! %or the &ovrite.
Newsom 3
USP 656 Multilevel Regression
Winter 2006
Summary a!le
Random vs"
Fixed
#efinition Example $se in Multilevel
Regression
Varia!les Random variable6 ("# is ssume! to 'e mesure! with
mesurement error. ,he s&ores re %un&tion o% true
s&ore n! rn!om error; (2# the vlues &ome %rom n!
re inten!e! to generli2e to mu&h lrger )o)ultion
o% )ossi'le vlues with &ertin )ro''ilit( !istri'ution
(e.g., norml !istri'ution#; (<# the num'er o% vlues in
the stu!( is smll reltive to the vlues o% the vri'le s
it ))ers in the )o)ultion it is !rwn %rom. Fixed
variable6 ("# ssume! to 'e mesure! without
mesurement error; (2# !esire! generli2tion to
)o)ultion or other stu!ies is to the sme vlues; (<#
the vri'le use! in the stu!( &ontins ll or most o% the
vri'le7s vlues in the )o)ultion.
-t is im)ortnt to !istinguish 'etween vri'le tht is
varying n! vri'le tht is random. * %i+e! vri'le
&n hve !i%%erent vlues, it is not ne&essril( invrint
(e5ul# &ross grou)s.
Random variable:
)hotogr)hs
re)resenting in!ivi!uls
with !i%%ering levels o%
ttr&tiveness
mni)ulte! in n
e+)eriment, &ensus
tr&ks
Fixed variable:
gen!er, r&e, or
intervention vs. &ontrol
grou).
Pre!i&tor
vri'les in M4R
generll(
ssume! to 'e
%i+e!
Effects Random effect6 ("# !i%%erent sttisti&l mo!el o%
regression or *N89* mo!el whi&h ssumes tht n
in!e)en!ent vri'le is rn!om; (2# generll( use! i%
the levels o% the in!e)en!ent vri'le re thought to 'e
smll su'set o% the )ossi'le vlues whi&h one wishes
to generli2e to; (<# will )ro''l( )ro!u&e lrger
stn!r! errors (less )ower%ul#. Fixed effect6 ("#
sttisti&l mo!el t()i&ll( use! in regression n! *N89*
ssuming in!e)en!ent vri'le is %i+e!; (2#
generli2tion o% the results ))l( to similr vlues o%
in!e)en!ent vri'le in the )o)ultion or in other
stu!ies; (<# will )ro''l( )ro!u&e smller stn!r!
errors (more )ower%ul#.
Random effect6
rn!om e%%e&ts *N89*,
rn!om e%%e&ts
regression
Fixed effect6 %i+e!
e%%e&ts *N89*, %i+e!
e%%e&ts regression
-nter&e)t onl(
mo!els in M4R
re e5uivlent to
rn!om e%%e&ts
*N89* or
*N:89*.
Coefficients Random coefficient: term ))lies onl( to M4R
nl(ses in whi&h inter&e)ts, slo)es, n! vrin&es &n
'e ssume! to 'e rn!om. M4R nl(ses most
t()i&ll( ssume rn!om &oe%%i&ients. 8ne &n
&on&e)tuli2e the &oe%%i&ients o'tine! %rom the level."
regressions s t()e o% rn!om vri'le whi&h &omes
%rom n! generli2es to !istri'ution o% )ossi'le vlues.
=rou)s re &on&eive! o% s su'set o% the )ossi'le
grou)s.
Fixed coefficient: &oe%%i&ient &n 'e %i+e! to 'e non.
vr(ing (invrint# &ross grou)s '( setting its 'etween
grou) vrin&e to 2ero.
Rn!om &oe%%i&ients must 'e vri'le &ross grou)s.
:on&e)tull(, %i+e! &oe%%i&ients m( 'e invrint or
vr(ing &ross grou)s.
Random coefficient6
the level.2 )re!i&tor,
verge in&ome, is
use! to )re!i&t s&hool
)er%ormn&e in e&h
s&hool. -nter&e)t
vlues %or s&hool
)er%ormn&e re
ssume! to 'e
sm)le o% the inter&e)ts
%rom lrger )o)ultion
o% s&hools.
Fixed coefficient6
slo)es or inter&e)ts
&onstrine! to 'e e5ul
over !i%%erent s&hools.
/oth use! in
M4R. Slo)es n!
inter&e)t vlues
&n 'e
&onsi!ere! to 'e
%i+e! or rn!om,
!e)en!ing on
reser&hers
ssum)tions n!
how the mo!el is
s)e&i%ie!. ,he
vrin&e o% the
slo)es or
inter&e)ts re
&onsi!ere!
rn!om
&oe%%i&ients.