You are on page 1of 2

633 Correlation causal-comparative quasi-experimental experimental designs Note: Quotations from Burke Johnson's (2000) article "An important

ant difference bet een causal!comparati"e and correlational research is that causal! comparati"e studies in"ol"e t o or more groups and one independent "ariable# hile correlational studies in"ol"e t o or more variables and one $roup%" (&a' ( Airasian# 2000# )*+)% ,ausal!comparati"e educational research attempts to identify a causative relationship bet een an independent "ariable and a dependent "ariable% However, this relationship is more suggestive than proven as the researcher does not have complete control over the independent variable% -f the researcher had control o"er the independent "ariable# then the research would be classified as true experimental research% ! http:.. %mnstate%edu. asson.ed*0).ed*0)lesson/2%htm 0esearch desi$ns fall into t o broad classes: 1uasi!e2perimental and e2perimental% 32perimental studies are characteri4ed b' the abilit' to randomi4e sub5ects into treatment and control $roups% 6his randomi4ation $oes a lon$ a' to ard controllin$ for "ariables hich are not included e2plicitl' in the stud'% Because comparison $roups are not true randomi4ed control $roups in 1uasi!e2perimental studies# this t'pe of stud' has to control for confoundin$ "ariables e2plicitl' throu$h statistical techni1ues% For this reason, quasi-experimental studies are sometimes labeled correlational designs% ! http:..facult'%chass%ncsu%edu.$arson.7A8*9.desi$n%htm :raenkel# the first author of the leadin$ sellin$ educational research te2t (:raenkel (;allen# /<<*)% ans ers the 1uestion# ";h' do educational researchers %%% seem to belie"e that e"idence for cause and effect ill be an' stron$er in causal!comparati"e research than in correlational research=": ,ausal!comparati"e research in"ol"es comparin$ (thus the "comparati"e" aspect) t o $roups in order to e2plain e2istin$ differences bet een them on some "ariable or "ariables of interest% 6he onl' difference bet een causal!comparati"e and e2perimental research is that the $roups bein$ compared in causal!comparati"e research ha"e alread' been formed# and an' treatment (if there as a treatment) has alread' been applied% >f necessit'# the researcher must e2amine the records of the t o $roups to see if he or she can offer a reasonable e2planation for (i%e%# hat "caused") the e2istin$ differences bet een the t o $roups%%%,orrelational research# on the other hand# does not look at differences bet een $roups% 0ather# it looks for relationships ithin a sin$le $roup% 6his is a bi$ difference%%%one is onl' entitled to conclude that a relationship of some sort e2ists# not that "ariable A caused some "ariation in "ariable B%%%-n sum# causal-comparative research does allow one to make reasonable inferences about causation; correlational research does not. (emphasis added) Based on the abo"e 1uotations# and the space allocated to these t o methods in popular te2tbooks# it should not be surprisin$ that almost ?0 percent (n@))0) of the participants in an All'n and Bacon (:all# /<<*) sur"e' of teachers of educational research said that the distinction bet een causal!comparati"e research and correlation research should be retained% Althou$h the ma5orit' supported the distinction# onl' a minorit' of those repl'in$ to a follo !up open!ended 1uestion (askin$ h' the' felt the distinction as needed) indicated a clearl' fault' "ie of causalit' in none2perimental research% Nonetheless# it is a problem hen an' teachers of educational research ha"e a fault' "ie of causalit'A

Causal-Comparative designs roblem: Be$ree at hich a researcher can determine the causal!effect ith an' measure of certaint' A relationship stud' of C and D ma' re"eal that chan$es in C cause chan$es in D# D causes C# or another "ariable E causes both C and D% F-%e%# the "third "ariable" problem# hich affects all correlational!t'pe desi$ns%G Causal-comparative versus correlational designs Neither is experimental Neither involves manipulation of a treatment variable 0elationships are studied in both ,orrelational: focus on ma$nitude and direction of relationship ,ausal!,omparati"e: focus on difference bet een t o $roups ! "true" experiment includes several ke# features: H one or more control $roups H one or more e2perimental $roups H random allocation to control and e2perimental $roups $n a true experiment the investigator does two things% a) randoml' assi$ns participants to $roups (e%$%# e2perimental and control)I and b) manipulates at least one independent "ariable% ! quasi experiment is almost the same& except now there is no random assignment of participants to groups% In order to reach "cause and effect" conclusions about the effect of the independent variable on the dependent variable, you must use a true experiment. ! http:..sociolo$'inde2%com.trueJe2periment%htm 'uasi-experiment is a research desi$n ha"in$ some but not all of the characteristics of a true e2periment% 6he element most fre1uentl' missin$ is random assi$nment of sub5ects to the control and e2perimental conditions% FNote that this is the same as "causal!comparati"e%"G 'uasi-experiments ! facult'%nc c%edu.toconnor.)0?.)0?lect0*%htm 6he ord "1uasi" means as if or almost# so a 1uasi!e2periment means almost a true e2periment% 6here are man' "arieties of 1uasi!e2perimental research desi$ns# and there is $enerall' little loss of status or presti$e in doin$ a 1uasi!e2periment instead of a true e2periment# althou$h 'ou occasionall' run into someone ho is biased a$ainst 1uasi!e2periments% (ummar#% )Causal-comparative) and )quasi-experimental designs) are equivalent % ! No random assi$nment of participants to $roups ! >ften there is no manipulation of -K Fi%e%# the diffs bet een the $roups are pre!e2istin$%G ! "0andom assi$nment" is a re1uirement for makin$ causal statements% ! "Lanipulation of the -K" is a 2nd re1uirement for makin$ causal statements% ! Thus, both fail as "causal" designs# althou$h the' ma' be "er' useful ( informati"e sources of information about differences bet een $roups% ! ,ausal inferences based on either ould be entirel' con5ectural# and ould re1uire a follo !on true e2periment to determine hether the causal con5ecture is correct or not%

You might also like