1. variables and research design

Upload: farahis

Post on 07-Aug-2018

220 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/20/2019 1. Variables and Research Design

    1/30

    Variables and researchdesign

    Anela Hasanagic, Ph.D

  • 8/20/2019 1. Variables and Research Design

    2/30

    ‘There are three sorts of lies: lies, damned liesand statistics.’ Disraeli.

    Statistics give us information about factors that

    we can measure. In research the things that we measure are called

    variables.All things that we can measure and record and

    that vary from one situation or person to another.

    hy

    statistics!!!

  • 8/20/2019 1. Variables and Research Design

    3/30

    Variables -characteristicsContinuous variables can take on absolutely any value

    within a given range. "hose are the most precise measures#li$e %&.'()%*

    Discrete variables can only take certain discrete values in

    a range. "hese can only be recorded in terms of presence orabsence of symptoms or some other characteristic andtherefore in terms of whole symptoms present.

    Categorical variables are those in which we simplyallocate people to categories. good e+ample is gender, which

    has only two values that it can ta$e male or female.-ategorical variables can also sometimes have manypossible values, as in type of occupation #e.g. udges,teachers, miners, grocers, civil servants*. hen dealing withcategorical data we have an in/nite number of variables thatwe might wish to investigate.

  • 8/20/2019 1. Variables and Research Design

    4/30

  • 8/20/2019 1. Variables and Research Design

    5/30

    Which of the following arecontinuous, which are discrete

    and which are categorical?ind speed "ypes of degree o0ered by a university1evel of e+troversion2a$es of car3ootball teams4umber of chess pieces ‘captured5 in a

    chess gameeight of giant pandas4umber of paintings hanging in art

    galleries

  • 8/20/2019 1. Variables and Research Design

    6/30

    Dichotomising continuousand discrete variables

    It is often the case that researchers convert continuous or discrete variablesinto categorical variables. 3or e+ample, we might wish to compare thespatial ability of tall and short people. e could do this by comparing peoplewho are over 678 cm with those under 6%9 cm on a spatial ability test. "hus,we have chosen points on the continuous scale #height* and decided to

    compare those participants who score above and below these points.Another e+ample might be to compare the memory ability of an+ious and

    non:an+ious individuals. e could measure an+iety levels using a;uestionnaire< this is a continuous variable measured on a discrete scale. 3ore+ample, the Hospital An+iety and Depression Scale has an an+iety scalethat ranges from & to '6. "o convert this to a categorical variable we wouldsimply compare those who score above a certain value #say, 66* with thosewho score below this value.

     "his dichotomising of continuous and discrete variables is ;uite common inpsychology as it enables us to /nd out if there are di0erences betweengroups who may be at the e+tremes of the continuous or discrete variables#e.g. tall and short people*. =ut it is not recommended because it reducesthe sensitivity of statistical analyses.

  • 8/20/2019 1. Variables and Research Design

    7/30

    Levels of measurment

  • 8/20/2019 1. Variables and Research Design

    8/30

    Levels of measurment

     "here are four levels ofmeasurement and these vary as a

    function of the information eachgives about the variables. "he fourdi0erent levels are

    6. 4ominal'. >rdinal8. Interval%. ?atio.

  • 8/20/2019 1. Variables and Research Design

    9/30

    1. Nominal scales

    At the lowest level of measurement are nominal scales.These are in eect categorical variables in that theyrepresent di0erent categories, but they also have thecharacteristic that there is no particular order that can begiven to the categories.

    A good e+ample of a nominal scale is gender, which has twocategories, male and female. You should be able to see thatthere is no logical way of ordering these two categories interms of magnitude.

    Another e+ample would be ethnic group again we cancategorise people in terms of their ethnic group but wecould not put these groups in any particular order @ they aresimply di0erent categories.

    hen dealing with nominal:level measures, we are simply

    putting people into categories and the data we obtain are inthe form of frequency counts. requency counts simply tell

  • 8/20/2019 1. Variables and Research Design

    10/30

    1. Nominal Scales discrete

    4o numerical or ;uantitativeproperties

    -lassi/es the levels of a variable intocategoriesgroups

    Independent variables are oftennominal #a categorical variable ordiscrete*

    Properties Identity

  • 8/20/2019 1. Variables and Research Design

    11/30

    !. "rdinal scales

    !uite often in psychology we use ratings scales tomeasure participants5 responses.

    3or e+ample, we might want to $now how nervous aperson is ust before they ta$e part in a study we are

    running.Bsing such a scale we can place participants in some sort

    of order in terms of how nervous they are prior to thestudy. " would be able to say that someone who put acircle around the ‘65 was less nervous than someone who

    put a circle around the ‘85 or around the ‘(5. >ne of thedrawbac$s with such scales is that we cannot say that thedi0erence between ‘65 and ‘'5 on the scale is the same asthe di0erence between ‘85 and ‘%5

     "hus we do not have e;ual intervals on the scale.

  • 8/20/2019 1. Variables and Research Design

    12/30

    "rdinal Scales generall# discrete

    ?an$ orders the levels of thevariable

    Allows categories to be ordered /rstto last, highest to lowest, biggest tosmallest, etc.

    Cuantitative but no values attachedto the intervals

    Properties Identity and 2agnitude

  • 8/20/2019 1. Variables and Research Design

    13/30

    $. %nterval scales

    At the interval level of measurement, we are able to put scores in somesort of order of magnitude and we also have e;ual intervals betweenadacent points on the scale #hence interval scale#.

     $ good e%ample of an interval scale is one of the commonly used scales tomeasure temperature, such as -entigrade or 3ahrenheit. >n such scales

    we can say that the di0erence between 6 and ' degrees is the same asthe di0erence between 7 and 6& degrees or between 77 and 6&& degrees.e have e;ual intervals between adacent points on the scales.

     "he disadvantage of such scales is that there is no absolute ero on them. "hus whilst there are ero points on both the -entigrade and 3ahrenheitscales these are arbitrary ero points @ they do not e;uate to ero

    temperature. "he ero point on the -entigrade scale was chosen as it wasthe point at which water freees, and the ero point on the 3ahrenheitscale is e;ually arbitrary. hen we reach ero on these scales we cannotsay that there is no heat or no temperature. =ecause of this we cannot saythat % E- is half as warm as F E- or that %& E- is twice as hot as '& E-.

    In order to ma$e such statements we would need a measuring scale that

    had an absolute rather than an arbitrary ero point.

  • 8/20/2019 1. Variables and Research Design

    14/30

    %nterval Scales

    Di0erence between the numbers ismeaningful

    Intervals are e;ual in sieCuantitative but no meaningful ero

    reference pointProperties Identity and 2agnitude

    and G;ual unit sie

  • 8/20/2019 1. Variables and Research Design

    15/30

    &. 'atio scales

     "he /nal level of measurement is the ratio scale. &atioscales have all the features of interval:level data butwith the addition of having an absolute ero point.

    G+ample of a ratio scale is speed of a car. hen thecar is not moving it has ero speed #an absolute eropoint* and the di0erence between 7 and 6& $.p.h. isthe same as that between '7 and 8& $.p.h. "he usefulpoint about having an absolute ero is that we can

    form ratios using such scales #hence ratio scales#.Thus, we can say that a car moving at '(( k.p.h. is

    moving twice as fast as one moving at (& $.p.h. >r aperson who read this paragraph in 8& seconds read ittwice as fast as someone who read it in )& seconds.

  • 8/20/2019 1. Variables and Research Design

    16/30

    'atio Scales

    Cuantitative with all numericalproperties including an absolute ero

    reference pointProperties Identity2agnitudeG;ual unit sieAbsolute ero

  • 8/20/2019 1. Variables and Research Design

    17/30

    What are some a((ro(riatestatistics?

    Nominal %dentif# andclassif#

    Number of cases)*ode) *ode) +hiSuare +hi Suare

    >rdinal Gstablish ran$

    order

    2edian<

    PercentilesInterval 3ind distances or

    intervals betweenunits

    2ean< SD< etc.intervals between - lti A4>Aunits-orrelation< A4>A<

    t:test

    ?atio 3ind ratios,fractions

    Percent variability<same inferential testsas Interval

  • 8/20/2019 1. Variables and Research Design

    18/30

    'esearch designs

  • 8/20/2019 1. Variables and Research Design

    19/30

     "here are a number of di0erent statistical techni;uesthat we use to analye the data we have collected inresearch.

    >ne of the biggest factors in determining whichstatistical tests you can use to analye your data isthe way you have designed your study. "here areseveral ways to design a study and the way you do socan have a great inuence on the sorts of statistical

    procedure that are available to you. Sometimesresearchers wish to loo$ for di0erences between twogroups of participants on a particular variable and atother times they might want to see if two variables arerelated in some way.

  • 8/20/2019 1. Variables and Research Design

    20/30

    eam(les

     $n e%ample of looking for dierences between conditions might be the researchreported by )u*guen and +iccotti -((#. "n this study the researchers wereinterested in whether or not dogs facilitate social interactions and helpingbehaviours among adults. The researchers ran four dierent studies where maleand female researchers walked with and without dogs. "n two studies theresearcher approached people and asked for some money, in another study theresearcher dropped some coins to see if people would help to pick them up and in

    a /nal study a male researcher approached females in the street and asked themfor their phone numbers. "n each study the researcher did the tasks both with andwithout dogs.

    "n all four studies they found that helping behaviors were higher when theresearcher had a dog than when they didn’t have a dog.

     $n e%ample of research looking for relationships would be the research reported

    by erreira, 0artine1 and )uisande -((2#. "n this study the researchers wereinterested in the personality factors that might be related to risky driving behavior.They measured a number of factors including an%iety, independence, tough3

    mindedness, self3control and e%troversion. rom the variables measured, theyfound that only an%iety, self3control and independence were related to risky drivingbehavior.

    The statistical tests that we would use in these e%amples are called dierence tests

    and correlation tests respectively.

  • 8/20/2019 1. Variables and Research Design

    21/30

    traneous and confoundingvariables

    Extraneous variables are those variablesthat might have an impact on the othervariables that we are interested in but we

    may have failed to ta$e these into accountwhen designing our study.

     "he main reason for conducting researchunder laboratory conditions is to try to control

    e+traneous variables as much as possible./ confounding variable is a speci/c type of

    e%traneous variable that is related to both ofthe main variables that we are interested in.

  • 8/20/2019 1. Variables and Research Design

    22/30

    +orrelational designs

    If we wish to understand how and whycertain variables are related to each other,perhaps the simplest way to e+amine such

    relationships between variables is by use ofcorrelational designs. In such a design wemeasure the variables of interest and thensee how each variable changes in relation

    to the changes in the other variables.+orrelational designs are those that

    investigate relationships between variables.

  • 8/20/2019 1. Variables and Research Design

    23/30

    +ausation

    >ne of the important aims of science is toestablish what causes things to happen. In allbranches of science, researchers are trying todiscover causal relationships between

    variables. "o be able to do this more easily we need to be

    able to manipulate one variable #change itsystematically* and then see what e0ect this

    has on the other variables.>ne of the golden rules of correlational designsis that we cannot infer causation fromcorrelations.

  • 8/20/2019 1. Variables and Research Design

    24/30

    0he e(erimental design In order for us to establish causal relationships between variables more

    easily we need to manipulate one of the variables systematically andsee what e0ect it has on the other variables. Such a process isessentially that underta$en in e(erimental designs.

    >ne of the most widely used designs in science is the e+perimentaldesign, also called the true e(eriment.

     "he variable manipulated by the e+perimenter is called the

    independent variable #I* that is, its value is not dependent upon #isindependent of * the other variables being investigated. "he othervariable in such an e+periment is called the dependent variable #D*.

    It is called the dependent variable because it is assumed to bedependent upon the value of the I. Indeed, the purpose of thee+periment is to establish or dismiss such dependence.

    G+perimental designs are those where the e+perimenter manipulatesone variable called the independent variable #I* to see what e0ect thishas upon another variable called the dependent variable #D*. Ine+perimental designs we are usually loo$ing for di0erences betweenconditions of the I. A hallmar$ of e+perimental designs is randomallocation of participants to the conditions of the I.

  • 8/20/2019 1. Variables and Research Design

    25/30

    uasi-e(erimentaldesigns

    >ften in psychology we want to loo$ at variables that wecannot directly manipulate. If we want to compare malesand females in some way, we cannot manipulate the groupto which each participant belongs. e cannot randomlyallocate participants to the male and female conditions< theyare already either male or female.

    If you are ever unsure whether you are dealing with ane+perimental or a ;uasi:e+perimental design, then loo$ forrandom allocation of participants to conditions. If it is not afeature of the design, then you are most li$ely dealing with a;uasi:e+perimental design.

    Cuasi:e+perimental designs involve seeing if there aredi0erences on the dependent variable #D* betweenconditions of the independent variable #I*. Bnli$ee+perimental designs there is not random allocation ofparticipants to the various conditions of the I.

  • 8/20/2019 1. Variables and Research Design

    26/30

    Student activit#

     "he following is an e+tract from the abstract of a paperpublished by Stead, Shanahan and 4eufeld #'&6&*

    Procrastination and stress are associated with poorermental health, health problems, and treatment delay. e

    e+amine procrastination in the domain of mental health.Higher levels of procrastination and stress werepredicted to correlate with poorer mental health statusand fewer mental health help:see$ing behaviours.

    Bndergraduate participants #68( females, )( males*

    completed online ;uestionnaires on procrastination,stress mental health issues, and mental health help:see$ing behaviours.

    What sort of design is this stud# and what sort ofvariables are being measured?

  • 8/20/2019 1. Variables and Research Design

    27/30

  • 8/20/2019 1. Variables and Research Design

    28/30

    2etween-(artici(ants andwithin (artici(ants designs

    4ithin3participant designs have the sameparticipants in every condition of theindependent variable #I*. Gach participant

    performs under all conditions in the study.=etween:participant designs have di0erent

    groups of participants in each condition ofthe independent variable #I*. "hus, thegroup of participants in one condition ofthe I is di0erent from the participants inanother condition of the I.

  • 8/20/2019 1. Variables and Research Design

    29/30

    Within (artici(ants design

     "he main advantage of using within:participants designs is that you are able tocontrol for many inter:individual confounding variables. hen you use di0erentgroups of people in each condition, you run the ris$ of there being some variableother than your I that also distinguishes between your groups. Jou would, if thishappened, have a confounding variable. hen you use a within:participantsdesign, you have much greater control over such variables. =ecause you have the

    same people in all your conditions of the I, there will be much less confoundingvariation between your conditions. =y and large the same person will bring thesame problems or ;ualities to all conditions of your I.

    A second very attractive point about using within:participants designs is that youneed to /nd fewer participants in order to complete your research.

     "he problem here might be order eect. 5ne way to eliminate order eects is tointroduce counterbalancing into your design. "n counterbalancing you get one half

    of your participants to complete the /rst condition followed by the secondconditon.

    Another limitation of within:participants designs is that having participants ta$epart in both conditions means that they are more li$ely to realise the purpose ofthe e+periment. "his is a problem because participants usually want to do what thee+perimenter wants them to and so may perform how they believe they should dorather than how they would normally have done. "hese are called demand eects.

    i i

  • 8/20/2019 1. Variables and Research Design

    30/30

    2etween-(artici(antsdesigns>ne of the important features of between:participants designs

    is that because you have di0erent groups of participants ineach condition of the I, each participant is less li$ely to getbored, tired or frustrated with the study. As a result, they aremore li$ely to perform at an optimum level throughout. In a

    similar vein, your research is going to be less susceptible topractice e0ects and the participants are less li$ely to wor$ outthe rationale for the study.

    =etween:participants designs therefore reduce order anddemand e0ects, and you can, to a large e+tent, eliminatethese factors as e+traneous variables from your study.

    >n the negative side, you will need more participants than youwould for a completely within:participants design. Also, if youuse di0erent participants in each condition, you lose a certaindegree of control over the inter:participant confoundingvariables.