Newsom 1

USP 656 Multilevel Regression

Winter 2006
Vmv875875887f (i.e. 0 mg 5 mg !n" #0 mg$. %owever if t&e rese!r'&er w!nts to m!(e inferen'es
)e*on" t&e +!rti'ul!r v!lues of t&e 'g&"f"gfg"fgsfgs"fgs"fg"ffffffffff((((("fg&"g&)le use" in t&e stu"*
! r!n"om effe'ts mo"el is use". , 'ommon e-!m+le woul" )e t&e use of +u)li' !rt wor(s
re+resenting low mo"er!te !n" 'v)')')'v)'v)
Fixed and Random Coefficients in Multilevel Regression
.&e r!n"om vs. fi-e" "&g"fg"fg"fg"fg"fg"fg"f(!n" effe'ts is im+ort!nt in multilevel regression. /n
multilevel regression mo"els )ot& level0# 'v'v)'v)v') to v!r* r!n"oml* !'ross grou+s. 1e'!use of
t&e !ssum+tions !)out t&eir error "istri)utions we '!ll t&eir v!ri!n'es oo=var(U0j) !n" 11=var(U1j),
2r!n"om 'oeffi'ients.3
.&is me!ns t&!t we '!n t&in( !)out 0j !n" 1j !s !(in to t&e r!n"om variables / "es'ri)e"
!)ove. /nste!" of !ttem+ting to gener!li4e )e*on" t&e +!rti'ul!r v!lues of t&e in"e+en"ent v!ri!)le
we !re !ttem+ting to gener!li4e )e*on" t&e +!rti'ul!r grou+s in t&e stu"*. 5or inst!n'e we m!* &!ve
50 'om+!nies )ut we wis& to gener!li4e to ! l!rger universe of 'om+!nies w&en we e-!mine t&e
me!ns (inter'e+ts$ or t&e -0* rel!tions&i+ (slo+es$.
/n %6M v!ri!n'es for t&e inter'e+ts !n" slo+es !re estim!te" )* "ef!ult. .&!t is w&en *ou
first 'onstru't ! mo"el U0j !n" U1j !re estim!te" )* "ef!ult. %owever t&e rese!r'&er &!s t&e '&oi'e
of setting U0j !n" U1j to )e 4ero. 1* setting t&em to 4ero we !re testing ! mo"el in w&i'& we !ssume
0j !n" 1j "o not v!r* r!n"oml* !'ross grou+s. /n f!'t t&eir v!ri!n'e is !ssume" to )e 4ero so t&e*
!re !ssume" to )e 'onst!nt or 2nonv!r*ing3 !'ross grou+s. 5or e-!m+le fi-e" nonv!r*ing inter'e+ts
woul" im+l* t&e grou+ !ver!ge for t&e "e+en"ent v!ri!)le is !ssume" to )e e7u!l in e!'& grou+.
,lt&oug& we t*+i'!ll* refer to t&is 'onstr!int !s 2fi-ing t&e inter'e+ts3 or 2fi-ing t&e slo+es3 t&e term is
somew&!t loosel* !++lie" )e'!use we !re re!ll* !ssuming t&e* !re fi-e" and nonvarying.
Intercept Model and Random Effects ANOVA
/n t&e sim+lest %6M mo"el t&ere !re no +re"i'tors. We &!ve one level0# e7u!tion !n" one level02
e7u!tion8
0 ij j ij
Y R = +
w&ere Y is t&e "e+en"ent v!ri!)le 0j is t&e inter'e+t !n" Rij is t&e resi"u!l or error. 0j '!n )e
t&oug&t of !s t&e me!n of e!'& grou+. .&e level02 e7u!tion !lso &!s no +re"i'tors in it9s sim+lest
form8
0 00 0 j j
U = +
w&ere
0j
is t&e "e+en"ent v!ri!)le
00
is t&e level02 inter'e+t !n" U
0j
is t&e level02 error. /n t&is
e7u!tion 00 re+resents t&e gr!n" me!n or t&e me!n of t&e inter'e+ts. U0j re+resents t&e "evi!tion of
e!'& me!n from t&e gr!n" me!n. W&en t&e !ver!ge "evi!tion is l!rge t&ere !re l!rge grou+
"ifferen'es.
/t '!n )e s&own t&!t t&is is re!ll* t&e ,N:V, mo"el. ,N:V, e-!mines t&e "evi!tions of grou+ me!ns
from t&e gr!n" me!n. 1e'!use we !ssume t&!t t&e grou+ me!ns re+resente" )* 0j !n" t&us t&eir
"evi!tions !re r!n"oml* v!r*ing t&is mo"el is e7uiv!lent to t&e random effects ,N:V, mo"el. We
'!n rewrite our two %6M e7u!tions !s ! single e7u!tion if we +lug in t&e level02 terms into level0#
e7u!tion.
00 0 ij j ij
Y U R = + +
/n t&is e7u!tion 00 re+resents t&e gr!n" me!n t&e U0j term re+resents t&e "evi!tions from t&e me!n
(or me!n "ifferen'es or tre!tment effe't if *ou woul" li(e$ !n" Rij re+resents t&e wit&in0grou+ error or
v!ri!tion. /n more m!t&em!ti'!ll* oriente" ,N:V, te-t)oo(s *ou will see !n ,N:V, mo"el written
somet&ing li(e8
Newsom 2
USP 656 Multilevel Regression
Winter 2006
ij j ij
Y e = + +
/f t&e mo"el is !n r!n"om effe'ts ,N:V, j is !ssume" to &!ve ! r!n"om "istri)ution. /n multilevel
regression )e'!use we !ssume U0j to v!r* r!n"oml* t&e sim+le %6M mo"el wit& no level0# or level02
+re"i'tors is e7uiv!lent to t&e r!n"om effe'ts ,N:V, mo"el.
W&en we !"" +re"i'tors to t&e level0# e7u!tion t&e* !re 'ov!ri!tes !n" t&e mo"el )e'omes !
r!n"om effe'ts ,N;:V, in w&i'& t&e me!ns !re !"<uste" for t&e 'ov!ri!te.
Newsom 3
USP 656 Multilevel Regression
Winter 2006
Summary a!le
Random vs"
Fixed
#efinition Example $se in Multilevel
Regression
Varia!les Random variable8 (#$ is !ssume" to )e me!sure" wit&
me!surement error. .&e s'ores !re ! fun'tion of ! true
s'ore !n" r!n"om error= (2$ t&e v!lues 'ome from !n"
!re inten"e" to gener!li4e to ! mu'& l!rger +o+ul!tion
of +ossi)le v!lues wit& ! 'ert!in +ro)!)ilit* "istri)ution
(e.g. norm!l "istri)ution$= (>$ t&e num)er of v!lues in
t&e stu"* is sm!ll rel!tive to t&e v!lues of t&e v!ri!)le !s
it !++e!rs in t&e +o+ul!tion it is "r!wn from. Fixed
variable8 (#$ !ssume" to )e me!sure" wit&out
me!surement error= (2$ "esire" gener!li4!tion to
+o+ul!tion or ot&er stu"ies is to t&e s!me v!lues= (>$
t&e v!ri!)le use" in t&e stu"* 'ont!ins !ll or most of t&e
v!ri!)le9s v!lues in t&e +o+ul!tion.
/t is im+ort!nt to "istinguis& )etween ! v!ri!)le t&!t is
varying !n" ! v!ri!)le t&!t is random. , fi-e" v!ri!)le
'!n &!ve "ifferent v!lues it is not ne'ess!ril* inv!ri!nt
(e7u!l$ !'ross grou+s.
Random variable:
+&otogr!+&s
re+resenting in"ivi"u!ls
wit& "iffering levels of
!ttr!'tiveness
m!ni+ul!te" in !n
e-+eriment 'ensus
tr!'(s
Fixed variable:
gen"er r!'e or
intervention vs. 'ontrol
grou+.
Pre"i'tor
v!ri!)les in M6R
gener!ll*
!ssume" to )e
fi-e"
Effects Random effect8 (#$ "ifferent st!tisti'!l mo"el of
regression or ,N:V, mo"el w&i'& !ssumes t&!t !n
in"e+en"ent v!ri!)le is r!n"om= (2$ gener!ll* use" if
t&e levels of t&e in"e+en"ent v!ri!)le !re t&oug&t to )e
! sm!ll su)set of t&e +ossi)le v!lues w&i'& one wis&es
to gener!li4e to= (>$ will +ro)!)l* +ro"u'e l!rger
st!n"!r" errors (less +owerful$. Fixed effect8 (#$
st!tisti'!l mo"el t*+i'!ll* use" in regression !n" ,N:V,
!ssuming in"e+en"ent v!ri!)le is fi-e"= (2$
gener!li4!tion of t&e results !++l* to simil!r v!lues of
in"e+en"ent v!ri!)le in t&e +o+ul!tion or in ot&er
stu"ies= (>$ will +ro)!)l* +ro"u'e sm!ller st!n"!r"
errors (more +owerful$.
Random effect8
r!n"om effe'ts ,N:V,
r!n"om effe'ts
regression
Fixed effect8 fi-e"
effe'ts ,N:V, fi-e"
effe'ts regression
/nter'e+t onl*
mo"els in M6R
!re e7uiv!lent to
r!n"om effe'ts
,N:V, or
,N;:V,.
Coefficients Random coefficient: term !++lies onl* to M6R
!n!l*ses in w&i'& inter'e+ts slo+es !n" v!ri!n'es '!n
)e !ssume" to )e r!n"om. M6R !n!l*ses most
t*+i'!ll* !ssume r!n"om 'oeffi'ients. :ne '!n
'on'e+tu!li4e t&e 'oeffi'ients o)t!ine" from t&e level0#
regressions !s ! t*+e of r!n"om v!ri!)le w&i'& 'omes
from !n" gener!li4es to ! "istri)ution of +ossi)le v!lues.
?rou+s !re 'on'eive" of !s ! su)set of t&e +ossi)le
grou+s.
Fixed coefficient: ! 'oeffi'ient '!n )e fi-e" to )e non0
v!r*ing (inv!ri!nt$ !'ross grou+s )* setting its )etween
grou+ v!ri!n'e to 4ero.
R!n"om 'oeffi'ients must )e v!ri!)le !'ross grou+s.
;on'e+tu!ll* fi-e" 'oeffi'ients m!* )e inv!ri!nt or
v!r*ing !'ross grou+s.
Random coefficient8
t&e level02 +re"i'tor
!ver!ge in'ome is
use" to +re"i't s'&ool
+erform!n'e in e!'&
s'&ool. /nter'e+t
v!lues for s'&ool
+erform!n'e !re
!ssume" to )e !
s!m+le of t&e inter'e+ts
from ! l!rger +o+ul!tion
of s'&ools.
Fixed coefficient8
slo+es or inter'e+ts
'onstr!ine" to )e e7u!l
over "ifferent s'&ools.
1ot& use" in
M6R. Slo+es !n"
inter'e+t v!lues
'!n )e
'onsi"ere" to )e
fi-e" or r!n"om
"e+en"ing on
rese!r'&ers
!ssum+tions !n"
&ow t&e mo"el is
s+e'ifie". .&e
v!ri!n'e of t&e
slo+es or
inter'e+ts !re
'onsi"ere"
r!n"om
'oeffi'ients.