Reasearch methods

Aim:

A general statement of what the researcher intends to investigate – purpose of investigation

Directional/one tailed hypothesis:

• States direction of the difference or relationship

• Uses words like ‘More, less, higher and lower’

Non-directional/two tailed hypothesis:

• Where direction of the differences is not stated.

• Simply states that there’s a difference, just not specific about the nature of the difference

• Eg, ppl who drink coffee differ in terms of talkativeness compared w/ ppl that don’t

Null hypothesis:

• States there’s no difference and IV has no effect on DV and if there’s a difference it’s due to chance

When to use which hypothesis

• Directional: When findings of previous research studies suggest a particular outcome- so used if previous research is in agreement

• Non-directional: Used when there’s no previous research, or findings from previous research are contradictory.

IV:

The variable manipulated/changed by the researcher and is the difference betw the 2 conditions

DV:

What the researcher measures. Any effect on the DV should be caused by changes in the IV.

Levels of the IV:

• In order to the test the effect of the IV we need different experimental conditions

• So 2 conditions of 2 levels of the IV are:

• Control and experimental condition

• (see effect of IV and compare)

Extraneous variable:

Any variable, other than IV, that may have an effect on the DV if not controlled -Doesn’t vary systematically w/ the IV.

Confounding variable:

Any variable, other than the IV, that may have affected the DV, so we can’t sure of the changes to DV

CV’s vary systematically w/ IV

Overcoming confounding variables

Randomisation: Use of chance to reduce researcher’s influence

Standardisation:

• Using the same procedures and instructions for all ppts in the study. Eg, same environment, information and experience

• Means that non-standardised changes in procedure don’t act as EVs.

Control groups and conditions

• Control groups used for purpose of comparison

• If change in behaviour of experimental group is significantly greater than control group, then researcher can conclude that the cause of this effect was the IV (assuming other CV’s constant)

Operationalisation

• Clearly defining variables in terms of how they can be measured

• Ensuring that variables are in a form that can be easily tested

Standardised procedures

• Set of procedures that are the same for all ppts in order to be able to repeat the study

• Includes standardised instructions. Instructions given to ppts to tell them how to perform the task

• Imp to make sure that each ppt do the exact same things in each condition otherwise results may vary bcs of changes in procedure rather than bcs of IV

Demand characteristics (-ve of exps)

• A cue that makes ppts unconsciously aware of the aims of study

• Help ppts work out what the researcher expects to find

• May lead to a ppt changing their behaviour

• May act in a way that they think is expected and over-perform to please experimenter or deliberately under-perform to sabotage results of study.

Overcoming the effects of the CV of demand characteristics

• Single blind technique- ppts hidden from purpose of research

• Double blind technique-  collection of data carried by another person (3^rd party). Done so that neither the ppts or experimenter I aware of what results or aim are expected or can influence/confound research

• Other details kept hidden, eg, which condition they’re in or if there’s another condition

Investigator effects (CV/EV)

• Any effect of the investigator’s behaviour, conscious or unconscious.

• Possible effects on design of study, selection of study, collection of data.

• These effects could be caused by tone of voice, age, race, gender etc… eg, leading q’s

Attempt to control investigator effects:

Randomisation

• Use of chance in order to control the effect of bias when designing materials and deciding the order of conditions (design of exp)

Reliability:

Refers to consistency of a research study

Validity:

Refers to whether an observed effect is a genuine one

Internal validity:

• Used to establish whether IV has affected DV and caused the results in an experiment- or did something else affect DV, a CV?

• Whether the researcher tested what they intended to test

• Whether study possessed/lacked mundane realism

• To gain high internal validity, CV’s need to be controlled and researcher is testing what they intend to test

External validity  (3 types)

• Considers the extent to which findings of  study can be generalised to a larger population and whether it can be applies to everyday life

• Affected by internal validity, can’t generalise low internal validity bcs results would have no real meaning for behaviour in question

Ecological validity:

• May not be appropriate to generalise from research setting (artificial) to other settings like everyday life

Population validity:

• Is sample in study representative of society as whole?

• Eg, if only men used, not appropriate to generalise findings to all ppl

Historical validity (temporal):

• The historical period, eg, if a study conducted in the 50’s, may not be appropriate to generalise findings to ppl today

Mundane realism

Refers to how a study mirrors the real world and how realistic the research environment is, to the degree to which experiences encountered in research environment will occur in the real world

Experimental designs:

Refers to how psychologists allocate people to the different conditions in an exp.

3 types:

Repeated measures, matched pairs, independent groups

RMD

Where all ppts experience both conditions of the experiment. From this, the 2 sets of data from both conditions would be compared to see if there’s a difference

+ves

• Ppts variables controlled, as same ppts used

• Fewer ppts needed since same ppts used in both conditions

-ves

• Order effects – order of conditions may affect performance, eg, ppts may do better on 2^nd test bcs of practice or less anxious as they’ve done it before. OR may do worse on 2^nd test bcs bored w/ doing same test again.

• Demand characteristics- as more likely ppts figure out aim of study as experienced all conditions and this may affect their behaviour

Dealing w/ the -ves of RMD

• Use 2 different tests to reduce practice effects

• Counterbalancing: to overcome order effects – ensures each condition tested 1^st/2^nd in equal amounts

AB/BA:

• Divide ppts into 2 groups

• G1: each ppt does A then B

• G2: B then A

• Still RMD even though 2 groups of ppt, bcs ppts do both conditions

ABBA

• All ppts take part in each condition twice

  Trial 1 : Condition A

  Trial 2 : Condition B

  Trial 3 : Condition B

  Trail 4 : Condition A

• Then compare scores on T1 and T4 w/ T2 and T3. Still RMD as comparing scores of the same person

IGD

• When 2 separate groups of ppts experience 2 different conditions of the experiment.
• Each group experiences one level of IV- either experimental or control condition.
• Performance of the 2 groups would then be compared

+ves

• Order effects not a problem as one group does one condition, other group 2^nd condition, whereas they’re a problem for RMD and ppts less likely to guess the aim

-ves

• Ppts in each condition are different. Researcher can’t control the effect of ppt variables (individual differences). Eg, ppts in A have better memory than B- acts as a CV

• Less economical than RMD as needs more ppts tan RMD in order to end up w/ same amount of data

Dealing w/ -ves of IGD

• Random allocation: Would randomly allocate ppts to conditions which distributes ppt variables evenly

MPD

• Using 2 groups of ppts, but match ppts on key characteristics believed to affect performance on the DV. (IQ, chattiest).
• One member of pair allocated to group A and other to group B (then carries on as IGD).
• This done as an attempt to control confounding variable of participant variables. Usually needs pre-test for matching to be effective

+ves

• Ppts only take part in a single condition, so order effects and demand characteristics less of a problem

• Controls individual differences between ppts. Means no different people in different conditions, so we can compare their results

-ves

• Not possible to control all ppts variables bcs you can only match on variables known to be relevant but there’ll still be imp differences between them that may affect DV. Eg, match on memory abilities, but later find out that ppts had been taught memory strategies and should’ve matched on this

• Time consuming, expensive and difficult to match ppts on key variables, especially if pre-test is required - so less economical than other designs

Dealing w/ -ves MPD

• Restrict # of variables to match on, to make it easier

• Conduct a pilot study to consider key variables that might be important when matching

Pilot studies

• A small-scale trial run of a study to test any aspects of the design or procedure, w/ a view of making improvements

Aim of pilot studies

• Allows researcher to identify any potential issues and to modify the design or procedure, saving time and money in the long run

• Can see what needs to be adjusted w/o having invested large amounts of time, money in full scale study

• Researcher not interested in results – Used to check procedure in exps, q’s understandable in self-report methods, checking coding systems in observational studies.

Conducted in a highly controlled environment within which the researcher, manipulates the IV and records the effect on the DV, whilst maintaining control of EV’s

+ves

• Have high control over EV’s. Means researcher can ensure that any effect on DV likely to be a result of manipulation of IV. ∴can be more certain about establishing cause and effect- high internal validity

• Replication very possible bcs of high level of control. Ensures that new EV’s not introduced when repeating an experiment. Replication vital to see if finding are valid

-ves

• Lacks ecological validity. So ∴ lack generalisability. Lab environment artificial and unlike everyday life. In unfamiliar contexts ppts may behave in unusual ways so their behaviour can’t be generalised beyond research setting – low external validity

• Demand characteristics – As ppts aware they’re being tested in lab exp and so might change their behaviour

• Low mundane realism – tasks ppts asked to carry out in a lab exp may not represent real-life experience like recalling random words

IV is manipulated in a natural, more everyday setting

+ves

• Ppts unaware they’re participating in exp, so more natural behaviour

• Higher mundane realism than lab exps bcs environment more natural, so field experiments produce more valid behaviour. High external validity.

-ves

• More difficult to control EVs bcs in natural setting, so makes it harder to establish cause and effect between IV and DV and replication not often possible

• IV in field exp may lack realism ∴ field exp not more like everyday life than lab exps

• Ethical issues – if ppts unaware they’re being studied. Can’t consent to being studied and difficult to debrief them and manipulating and recording behaviour – invasion of privacy

Have an IV that’s naturally occurring, but IV hasn’t been made to vary by anyone.Simply a difference betw ppl that exists. Eg, gender, personality

+ve

·         Allows comparisons betw types of ppl

·         Often carried out in controlled conditions, so ∴ share +ves of lab: Highly controlled, E/CV reduced ∴ higher internal validity, can be easily replicated ∴ external validity

-ves

·         Can’t randomly allocate ppts to conditions and ∴ there may be CV’s

·         The unique characteristics of sample means findings can’t be generalised to other groups of ppl – low population validity

·         As IV not manipulated, may not be case that IV has effect on the DV

·         Can only be used where conditions vary naturally

·         Ppts may be aware of being studied so change behaviour∴  reducing internal validity

Population

The group of ppl that the researcher is interested in, from whom a sample is drawn, from whom generalisations can be made

Sample

• A group of ppl who take part in a research investigation

• Sample is drawn from a target population

• More representative sample is of target pop, the more the researcher can generalise

 5 sampling methods ROSSV

• Random

• Opportunity

• Stratified

• Systematic

Random sampling

• Where every member of the target pop being tested have an equal chance of being selected

• 2 ways: lottery method and random generator

• Lottery: compile list of all ppl in pop, put names in hat and select # of names required

• Random generator: # every member of pop, using computer random generator, generate #s

+ves

• Free from researcher bias. Researcher has no influence over who’s selected and so prevents them from choosing ppl who they think may support their hypothesis and as all members of the pop have an equal chance of being selected ∴ good representation of target pop

-ves

• Difficult and time consuming to conduct as you need a list of all members of the population, then need to contact selected…

• May end up w/ sample that’s still unrepresentative. Possible to select similar ppts, eg, all females

• Selected ppts may refuse to take part, which means researcher will be left w/ a volunteer sample (helpful ppl) and all limitations of that

Systematic sampling

• When ppts chosen in a systematic way. A sampling frame produced- this is a list of ppl in the target pop organised into for eg, alphabetical order, door no’s.
• Then every nth ppt (5^th,7^th) on list of names from the sampling frame is chosen- or randomly to avoid bias

+ves

• Avoids researcher bias. Once sample frame has been established, researcher has no influence over who’s chosen – as ppts chosen using an objective system

• Fairly representative

-ves

• Periodic traits in population. Eg, every 5^th door has female living ∴ sample would be unrepresentative

• Not truly unbiased unless you select a # using s random method and start w/ this person and then select every nth person

Stratified sampling

A sample of ppts produced by identifying subgroups according to their frequency in population. Ppts then selected randomly from subgroups

How it’s done:

• First researcher identifies the different strata that make up the population

• Then, the proportions needed for the sample to be representative are worked out

• Finally, the ppts that make up each stratum are selected using random sampling

+ves

• Avoids researcher bias. Once the target pop has been sub-divided into strata, the ppts that make up the No’s are randomly selected and so beyond influence of researcher

• Representative sample produced bcs it’s designed to accurately reflect the composition of the pop – means generalisation of findings are possible

-ves

• Very expensive and time consuming to identify subgroups, then randomly select ppts and contact them

• The identified strata can’t reflect all the possible ways ppl are different

• A sample of ppts produced by selecting ppl who are most easily available at the time of the study

• Eg, ppl walking by you in street or students at school

+ves

• Convenient – saves time and effort and much less costly in terms of time and money than random sampling

-ves

• Sample is unrepresentative of target pop as it’s drawn from a very specific area, like one street in one town, so findings can’t be generalised to target pop

• Researcher bias as researcher has complete control over the selection of ppts and for eg, may avoid ppl they don’t like the look of

Volunteer sample

• Involves ppts selecting themselves to be a part of the sample- self selection

• Done by researcher advertising study like in a newspaper

+ves

• Collecting s volunteer sample is easy. Requires minimal input from researcher as they come to you and so less time consuming than other sampling methods

-ves

• Volunteer bias- Asking for volunteers may attract a certain ‘profile’ of person- one that’s helpful. This may affect how far findings can be generalised

Gender bias-

• only males = androcentric
• only females = gynocentric

Cultural bias

• if done in one country than ethnocentric

• Naturalistic and controlled

• Covert and overt

• Participant and non-participant

• Behaviour studied in a natural situation where everything has been left as it is normally

• All aspects of environment are free to vary

+ves

• High external validity as findings can often be generalised to everyday life as behaviour is studied within the environment where it would normally have occurred

-ves

• The lack of control over the research situation makes replication of study difficult and may mean that something unknown to the observer may account for behaviour observed

• May be many uncontrolled extraneous variables that make it difficult to judge any pattern of behaviour

• Some variables in the environment are regulated by researcher reducing ‘naturalness’ of environment and behaviour being studied

• Ppts likely to know they’re being studied and may be conducted in lab

• Allows researcher to investigate effects of certain things on behaviour

• Some control over variables, including manipulating variables to observe effects and also control of extraneous variables

+ves

• Extraneous variables may be less of a factor, so replication of the observation becomes easier

• Observer can focus on particular aspects of behaviour

-ves

• May produce findings that can’t be as readily applied to real-life settings- as it lacks ecological validity

• Can cause demand characteristics, if ppts know they’re being observed

Covert observations

Ppts behaviour is watched and recorded w/o their knowledge and consent

+ves

• Ppts don’t know they’re being watched. This removes the problem of participant reactivity and ensures any behaviour observed will be natural. Increases validity of data

-ves

• Ethical issues- no informed consent or right to withdraw for ppts

 Overt observations

When ppts know their behaviour is being observed

+ves:

• More ethically acceptable as informed consent is obtained

-ves:

• Increases ppt reactivity (DC), making results less valid

Researcher becomes a member of the group whose behaviour they’re observing

+ves:

• Researcher experiences situations as ppts do, giving insight into ppl being studies- incs validity of findings

• Rich, detailed account of behaviour

-ves:

• Risk that observer may come to identify too strongly w/ those they’re studying and lose objectivity

• Problems of reliability due to observer bias

• Observer’s presence in group changes their behaviour- alters dynamics of group

• If overt- DC

• If covert- ethical issues

 Non-ppt observations

Researcher remains part of group they’re observing

+ves:

• Less ethical issues like deception as ppts know they’re being observed

• Allows researcher to maintain an objective distance from their ppts, so less likely to lose objectivity

-ves:

• Lose valuable insight into the behaviour of the ppl being observed

• May make results less valid as ppts may change their behaviour as they know R is present- DC

• Where researcher records data by simply writng down everything they see

• Used when obsv w/ few ppts

• Can be used as pilot study to see which behavioural categories to use

+ves:

• Data collected has depth and is rich

-ves:

• Risk of observer bias bcs objective behavioural categories not present (only 4 structural)

• Researcher may only record behaviours that catch their eye and these may not be the most important or relevant behaviours

• Too much to record

• Tends to produce qualitativ4e data- difficult to record and analyse

Structured observations

When researcher uses behavioural categories and sampling procedures to organise observations

+ves:

• Use of behavioural categories makes recording of data easier and more systematic

• Numerical data probs produced- quantitative, can analyse and compare behavs betw ppts more easier

-ves:

• Lack of depth in data collected

Behavioural categories

When a target behaviour is broken up into components that are observable and measurable

Behavioural categories should be:

• Objective- observer shouldn’t have to make inferences about the behav

• Cover all possible categories of behav

• Mutually exclusive- shouldn’t have to mark to categories at once

• Must be observable, measurable and clearly defined

Sampling methods

• A method to select which behavs to record

• Used bcs too much data to record, so it’s a systematic way of recording data

Event sampling

• Counting the # of times a certain behave occurs in target person/group

• Useful when target behavs occurs not often and would be missed if other sampling method used

Time sampling

• Recording behavs in a given time frame, noting what someone does every 30 secs

• Effective in reducing the # of observations that have to be made

• But, those periods, when behave is sampled, might be unrepresentative of the observation as a whole

Point sampling

• Observing and recording behaviour which occurs at a series of given points in time, eg, meal times

IOR: Observations carried out by at least 2 researchers

• Observers should familiarise w/ the behavioural categories to be used

• Then observe same behaviour at same itme

• Observers should compare data recorded and discuss any differences in interpretation, then should analyse data from study

Evaluation

• Allows access to ppl’s thoughts and feelings

• Social desirability bias

• Some don’t know what they think/feel so their answer lacks validity

• Sample used may not be representative and so data collected can’t be generalised

Questionnaires

Set of written questions used to assess thoughts and feelings

+ves:

• Cost effective- can be distributed to large # of ppl cheaply and quickly and can be completed w/o R

• Respondents may feel more willing to reveal personal info than in an interview

• Data that questionnaires produce easy to analyse, so comparisons can be made

-ves:

• Social desirability- answers not truthful, exaggerated and so invalid

• Acquiescence bias- tendency to agree w/ item regardless of content of question, so invalid

Structured interviews

Interviews that are made of a pre-determines set of Q’s that are asked in a fixed order

+ves:

• Easy to replicate bcs of their standardised format, so can confirm reliability. Also, answers form diff ppl can be compared and so easier to analyse than unstructured bcs answers more predictable

-ves:

• Not possible for interviews to deviate from topic or elaborate their points

• Interviewer bias- interviewer’s expectations may influence the answers the interviewee gives

Semi-structured interviews

• Composed of structural and unstructured bits

• List of Q’s made in advance, but interviewer free to ask more Q’s when they feel appropriate

Unstructured interviews

No set Q’s. There’s a general aim that a certain topic’ll be discussed. Interviewee is encouraged to expand and elaborate their answers

+ves

• More detailed info obtained bcs interviewer tailors the Q’s to the specific responses and can get deeper insights into thoughts and feelings

-ves:

• Respondents may lie/exaggerate bcs social desirability

• Analysis of data not easy- R has to sift through lots of irrelevant info and draw conclusions

• Requires interviewers w/ more skill bcs has to develop Q’s on spot, more expensive

Factors of good questions

• Clarity- if Q’s misinterpreted/confusing means info collected won’t be valid

• Jargon

• Emotive lang- Sometimes author’s attitude towards a particular topic clear from the way Q is worded

• Leading Q’s- Guides respondents towards a particular answer

• Double-barrelled- contains 2 Q’s in one- respondents may agree w/ one half of Q and not the other

• Double negatives- hard for respondent to decipher

• Analysis- Using closed Q’s

Things to consider when designing a questionnaire

• Filler Q’s- may help to include irrelevant Q’s to distract respondents from main purpose of study, reduce DC

• Sequence of Q’s- start w/ easy ones, so save Q’s that might make someone feel anxious until respondent’s relaxed

• Sampling technique- selected respondents must be representative to generalise findings

• Pilot study- Q’s can be tested on small groups of ppl. Means Q’s can later be improved in response to any flaws

Open Q’s:

Doesn’t have fixed range of answers and respondents free to answer in any way

Eval

• Respondents can expand on their answers- incs amount of detailed info collected

• Can provide unexpected answers, allowing Rs to gain new insights into ppl’s feelings and attitudes

• Respondents less literate may find OQ’s difficult and most respondents avoid long, complex answers ∴OQ’s may not provide detailed extra info

• Produces qualitative data-  hard to summarise bcs wide range of answers and difficult to identify trends

Closed Q’s:

Offers a fixed # of responses

Eval

• Have limited range of answers and produces quantitative data. Makes it easier to analyse using graphs and mean

• Respondents may be forced to select answers that don’t represent their real thoughts and feelings. Data collected lacks validity

• Acquiescence bias ∴ data collected not valid/informative

Designing interviews

• Q’s should be standardised for each ppt to reduce interviewer bias

• Start w/ neutral Q’s, so ppts relaxed and can establish rapport

Correlations

Shows the strength and direction an association betw 2 or more co-variables.

+ve: Ice cream sales and temperature

-ves: Age and beauty

0: no RS betw co-variables. Weather and subjects

Correlational hypothesis

States the expected association betw the co-variables

Correlation coefficient

A # betw -1 and 1.

+1 perfect +ve correlation

Difference betw correlations and experiments

• In an experiment, the R manipulates the IV in order to measure the effect on the DV. Bcs of this manipulation, it’s possible to infer that the IV caused the change in DV

• In correlation, there’s no manipulation of one variable and ∴not possible to establish cause and effect betw one co-variable and another

+ves of correlations

• Procedures in correlation can be easily repeated, means findings can be confirmed

• Correlations used to investigate trends in data. If correlation significant then further investigation is justified, if not then casual RS

• Quick and economical. No need for controlled environment and no manipulation of variables is required. Secondary data can be used- means less time consuming than exps

-ves of correlations

• Tells us how variables are related but not why. Bcs lack of experimental and control within a correlation. So can’t determine cause and ∴effect don’t know which co-variable causes the other to change

• Could be intervening variable causing the RS betw the 2 co-variables. Bcs of this, correlations can be misinterpreted, as RS betw variables presented as facts. Such misinterpretations of correlations may mean that ppl design programmes for improvement based on fake premises. Eg, single mum commit crime, may be bcs they’re less well off

Qualitative data

Thoughts and feelings expressed in words. Non-numerical

Evaluation

• Provides rich and detailed info. Allows R’s to develop their thoughts, feelings and ops

• Difficult to analyse and draw conclusions

Quantitative data

Data that’s expressed numerically, scores

Evaluation

• Simple to analyse, ∴comparisons betw groups easily drawn

• Numerical data tends to be more objective and less open to bias

• Such data fails to represent real life. Eg, closed Q’s may force ppl to tick answers that don’t represent their feelings∴  conclusions may be meaningless

Primary data

Refers to OG data that’s been collected specifically for purpose of investigation by R

Evaluation

• R has control over the data. Data collection can be designed so it fits the aims and hypotheses of the study

• V lengthy and ∴ expensive process. Designing a study takes time and the time spend recruiting ppts, conducting the study and analysing the data. V lengthy compared to secondary, where data can be accessed quickly

Secondary data

Data that’s been collected by someone other than the person who’s conducting the research.

Evaluation

• Requires minimal effort to access someone else’s data and cheaper bcs less time and equipment needed.

• Such data may’ve already been subjected to statistical testing ∴ whether significant or not is known

• Data may not fit needs of the study

• Data may be outdated/incomplete