AN I NT RODUCT I ON T O L AT E NT CL ASS ANAL YSI S F OR PRE VE NT I ON RE SE ARCH Be tha ny C. Bra y, Ph.D. T he Me tho do lo g y Ce nte r, Pe nn Sta te me tho do lo g y.psu.e du
OVE RVI E W • Co nc e ptua l intro duc tio n to la te nt c la ss a na lysis (L CA) • An e xa mple : L a te nt c la sse s o f a do le sc e nt drinking b e ha vio r • Pa ra me te rs e stima te d in L CA • T e c hnic a l c o nside ra tio ns: Mo de l ide ntific a tio n, mo de l se le c tio n • So ftwa re o ptio ns
OVE RVI E W • I nc luding g ro uping va ria b le s • Pre dic ting la te nt c la ss me mb e rship • Pre dic ting a dista l o utc o me • Que stio n & Answe r se ssio n: Ope n disc ussio n
ABBRE VI AT I ONS • L CA = la te nt c la ss a na lysis • Sta tic , c a te g o ric a l la te nt va ria b le • L T A = la te nt tra nsitio n a na lysis • Dyna mic , c a te g o ric a l la te nt va ria b le
ON: CA L I RODUCT NT I UAL PT CONCE
T HE BASI C I DE AS • I ndividua ls c a n b e divide d into sub g ro ups b a se d o n uno b se rva b le c o nstruc t • T he c o nstruc t o f inte re st is the la te nt va ria b le • Sub g ro ups a re c a lle d la te nt c la sse s
T HE BASI C I DE AS • I ndividua ls c a n b e divide d into sub g ro ups b a se d o n uno b se rva b le c o nstruc t • T he c o nstruc t o f inte re st is the la te nt va ria b le • Sub g ro ups a re c a lle d la te nt c la sse s • T rue c la ss me mb e rship is unkno wn • Unkno wn due to me a sure me nt e rro r • Me a sure me nt o f the c o nstruc t is typic a lly b a se d o n se ve ra l c a te g o ric a l indic a to rs
T HE BASI C I DE AS • I ndividua ls c a n b e divide d into sub g ro ups b a se d o n uno b se rva b le c o nstruc t • T he c o nstruc t o f inte re st is the la te nt va ria b le • Sub g ro ups a re c a lle d la te nt c la sse s • T rue c la ss me mb e rship is unkno wn • Unkno wn due to me a sure me nt e rro r • Me a sure me nt o f the c o nstruc t is typic a lly b a se d o n se ve ra l c a te g o ric a l indic a to rs • L a te nt c la sse s a re mutua lly e xc lusive & e xha ustive
T HE BASI C I DE AS • Ho w ma ny pe o ple a re fa milia r with fa c to r a na lysis? • Simila r to c o nfirma to ry fa c to r a na lysis b ut the la te nt va ria b le is c a te g o ric a l AND • Simila r to e xplo ra to ry fa c to r a na lysis b ut the la te nt va ria b le is c a te g o ric a l
E ST I MAT E D PARAME T E RS • L a te nt c la ss pre va le nc e s • e .g ., pro b a b ility o f me mb e rship in HI GH DE PRE SSI ON la te nt c la ss • I te m-re spo nse pro b a b ilitie s • e .g ., pro b a b ility o f re po rting F e lt L o ne ly g ive n me mb e rship in HI GH DE PRE SSI ON la te nt c la ss
AN E XAMPL E OF L CA De pr e ssion in adole sc e nc e ig ht indic a to rs (2 8 =256 po ssib le re spo nse pa tte rns) • E (L a nza , Bra y, & Co llins, 2013) Sad Dislike d • Co uldn’ t sha ke b lue s Pe o ple unfrie ndly • • F e lt de pre sse d Dislike d b y pe o ple • • F e lt lo ne ly F ailur e • F e lt sa d L ife wa s fa ilure • L ife no t wo rth living •
AN E XAMPL E OF L CA F ive late nt c lasse s of de pr e ssion L ate nt Class Pr e vale nc e No de pre ssio n 41% Sa d 18% Dislike d 17% Sa d & Dislike d 15% De pre sse d 9%
Conditional pr obabilitie s of AN E XAMPL E OF L CA ‘ye s’ r e sponse Cla ss 1 Cla ss 2 Cla ss 3 Cla ss 4 Cla ss 5 I ndic a to r No n-D Sa d Dislike d S+D De pre sse d Co uld no t .03 .54 .17 .66 .90 Sha ke Blue s F e lt De pre sse d .06 .73 .24 .86 1.00 F e lt L o ne ly .07 .58 .33 .77 .92 F e lt Sa d .14 .80 .38 .87 .94 Pe o ple .13 .17 .64 .67 .70 Unfrie ndly Pe o ple Dislike .04 .00 .68 1.00 .77 Yo u L ife wa s F a ilure .01 .11 .10 .23 .90 No t wo rth living .00 .06 .11 .10 .85
OR S OF HAVI ASSE NT NG BE SCE CL NT E I NK ADOL E AT DRI L
H GRADE N 12 T DRI NK I NG I • Da ta fro m 2004 c o ho rt o f Mo nito ring the F uture pub lic re le a se • N = 2490 hig h sc ho o l se nio rs who a nswe re d a t le a st o ne q ue stio n a b o ut a lc o ho l use (48% b o ys, 52% g irls) • Go a ls o f the study: Alc o ho l use b e ha vio r a mo ng U.S. 12 th g ra de rs • • Ge nde r diffe re nc e s in me a sure me nt a nd b e ha vio r • Pre dic t b e ha vio r fro m skipping sc ho o l a nd g ra de s
H GRADE N 12 T DRI NK I NG I Se ve n indic ator s of dr inking be havior I te m Pro po rtio n ‘ Ye s’ L ife time a lc o ho l use 82% Pa st-ye a r a lc o ho l use 73% Pa st-mo nth a lc o ho l use 50% L ife time drunke nne ss 57% Pa st-ye a r drunke nne ss 49% Pa st-mo nth drunke nne ss 29% 5+ drinks in pa st 2 we e ks 26%
Recommend
More recommend