Non Experimental Methods

Non Experimental Methods

Investigating Behaviour Non Experimental MethodsSurveysQuestionnaires written or verbalInterviewsStructured pre determined questions (questionnaire that is delivered face to face)Unstructured Each question is developed as a result of the previous answer. (Clinical interview)

Both can be used to collect quantitative or qualitative data.What ethical issues should be considered? (3)TaskWhat do you think the strengths and limitations would be for:Questionnaires (2strengths, 2 weaknesses)Structured interviews (3 and 3)Unstructured (2 and 3)

Work in pairs or 3s to discuss the development, use and analysis of data using these techniques to help you come up with your answers. Consider how you would feel if you were the participant.Compare and contrast them.EvaluationStrengthsWeaknessesQuestionnairesOnce questionnaire has been designed it is quick and easy to replicate to collect a large number of responses.Dont require specialists administrators.More willing to give information as feedback would be confidential rather than in interviews/Leading questions/social desirability bias - answers not truthful.Biased sample as only certain types of people are willing to fill in questionnaires.Structured InterviewsEasily repeated.Requires less skill than unstructured.Easier to analyse as answers are more structured/predictability.Interviewer bias.Data is restricted/limitedPeople feel less comfortable about revealing informationUnstructured InterviewsMore detailed information can be elicited.Information is less restrictedInterviewer bias (more likely than in structured)Requires well trained interviewers.Data may not be comparable (questions would differ)More difficult to analyse data/unpredictableGlossary of termsLeading questions a question which is worded in a way that makes one answer more liked. Wouldnt you agree . . .?Social Desirability Bias what people think they should say rather than genuinely think/feel.Interviewer Bias what the interviewee thinks the interviewer wants.

ReliabilityConsistency.Experimenters must ensure that the questionnaire they use is reliable to ensure that they are testing what they aim to.Ie If the same questionnaire is used with the same participant on 2 different occasions but gain different responses, if the questionnaire was reliable then we can assume that the difference can be attributed to the participant.Test-retest reliability repeating questionnaire with same participants to see if the same results are obtained and measured using a correlation coefficient.ValidityThat the questionnaire is measuring what it was designed to measure.If answers are affected by social desirability bias, interviewer bias etc then the results lack validity.Concurrent validity comparing results from new questionnaire with an established study. Good Questionnaires:Are clear.Lack biasEasy to analyse closed questions (options given, likert scale) but this makes the data quantitative rather than qualitative.

Good Surveys:Include filler questions misleads interviewee to reduce interviewer bias.Easy qstns at start more difficult at end.Pilot studyCorrelational Design and AnalysisCorrelational analysis is used to analyse non-experimental methods (surveys/observations/case studies) as the IV is not being manipulated and there is very little control over extraneous variables.

It is not a research method!Correlational Design and AnalysisCorrelation a relationship between two variables (co-variables).Positive correlation when 2 variables increase together.Negative correlation when 1 variable increases and the other decreases.

Zero CorrelationZero correlation no relationship between the 2 variables.

Visual DisplayCorrelations are displayed using a scattergram.A dot is plotted on a scattergram for each participants two answers. Eg weight and hours spent exercising per week.

Statistical TestIn the same way we used T-test for our experimental methods in order to reject our null hypothesis so too do we use statistics in non-experimental methods.Correlation co-efficient never greater than 1 (+1 for positive correltaions and -1 for negative correlations).1 Is the perfect correlation but is very rare. The strength of correlation is expressed as weak, moderate or strong.The closer to 1 the number is the stronger the correlation:+0.76 (there is a positive and strong correlation between the co-variables)-0.76 (there is a negative and strong correlation between the co-variables)+0.002 (There is a very weak positive correlation between the co-variables)SignificanceA table of significance is used to determine whether the strength of the correlation (the number) is significant.The significance is dependent upon the number of participants who took part/responded (N).Even small numbers (0.02) can be significant if the N is very large.Evaluation of investigations using correlational analysisStrengthsWeaknessesCan be used when it would be ethically wrong to manipulate variables.Even with a correlation it does not mean there is a cause-effect between the two variables. If the correlation is significant then further research should be carried out/justified.There may be other untested variables that could explain the correlation.If there is a zero correlation then causal relationships can be ruled out. (they dont cause each other find other variables)ObservationThe starting point for all investigations.Must be objective. If used as part of an experiment then it is a research technique. However, if used on its own it is a research method.

The design of these observations may vary:Naturalistic ObservationBehaviour is studied in a completely natural setting. The researcher does not change anything.

This is different from a natural experiment as the experimenters are not looking for a relationship between an IV and a DV.Example:Children aged 3- 5 observed in a playground. Activities were categorised as male, female or neutral. Praise and imitation was recorded as positive responses while criticism and stopping play were recorded as negative responses.

Children generally reinforced peers for gender-appropriate play and criticised gender-inappropriate play.

Lamb and Roopnarine, 1979Controlled ObservationSome variables can be controlled eg setting, objects etc. Participants are likely to know they are being studied.Reduces chances of behaviour being completely natural.ExampleLamb and Roopnarines experiment could also have been carried out as a controlled observation.ie The children could have been out in a playroom where the toys would have been pre chosen by the experimenters.

NB this is different from Banduras Bobo Doll experiment as they directly affected the IV (showing the children a video in advance). In this instance observation was used as a technique not a method.Other types of ObservationContent Analysis observing written/verbal texts.Non-participant observer does not involve themselves simply observes.Participant Observers participate.Disclosed participant knows they are being observed.Undisclosed participant unaware they are being observed.Example (page 119) Leon Festinger, a social psychologist, 1950s.Religious cults receiving messages from aliens detailing the end of the world.Believed they would be rescued by a UFO and arranged to meet a set time and location.Festinger posed as a convert in order to observe their reactions when their beliefs were unfounded.The night before the predicted flood, when it was apparent that there would be no flood, the cult leader claimed that it was because of their combined prayers.Observed a mixture of reactions some didnt believe and left the cult, others took it as evidence of the cults powers.ReliabilityTo ensure reliability (consistency) it is always better to have at least 2 observers who produce the same record (if 2 observers have conflicting recordings their observations are not reliable).Observers should be trained to increase level of reliability.The extent of this consistency is call inter-observer reliability Measured by correlating the results . To have inter-observer reliability the correlation must be >.80.Validity (measuring what its supposed to)Can be affected by observer bias the observer sees what they want/expect to see rather than being objective.http://www.youtube.com/watch?v=1GEEvvTiiQkUsing more than one observer and averaging their responses can help with this.If participants know they are being observed they may change their behaviour (eg social desirability bias).Results will not be valid if the coding system is flawed.

Task:In pairs decide who is A and who is B.Take it in turns to complete the task. Each task should take 5 minutes.While one person is completing the task the other person should be recording all aspects of their behaviour: ie facial expressions, verbal cues, body language, etcObservational techniquesStructured vs Unstructured.Unstructured (what you did): Observer records everything they believe to be relevant. No system for observing or recording behaviour.Observer will note only the most obvious behaviour but these may not be the most important (did you see the moonwalking bear?).

StructuredSystems are put in place to ensure observations are organised and systematic.Sampling procedure knowing who you are observing and when.Observation schedule how to record the behaviour you are studying.Continuous observation Every instance of the relevant behaviour is recorded in great detail. Difficult as it requires a lot of recording if the behaviour is common.Event sampling counting the amount of times it occurs.Time sampling recording behaviour at set intervals (every 30 seconds). Can be selected from a checklist.Observational Schedule/Coding system/Behaviour checklistHaving a pre-made list of relevant behaviour to be observed which helps the observer to deconstruct the behaviour.They should be objective and leave no room for observer bias.Be wide spread cover all possibilities.Should be discrete no room for overlapshttp://www.cs.cmu.edu/afs/cs/project/face/www/facs.htm

Research Methods - SummaryPart 1 - http://www.youtube.com/watch?v=Lkwsrh10OXM

Part 2 - http://www.youtube.com/watch?v=QwcuHaJ7q0ICase studiesA detailed study of an individual/place/event.How Freud collected his data Anna O.Requires a variety of research methods to develop an informed picture interviews, tests etc.Extremely in depth.Time consuming (longitudinal) carried out over a long period of time.

Task: What are the advantages and disadvantages of this method?Try to come up with 3 of each and use your prior knowledge of this topic to help.

EvaluationAdvantagesDisadvantagesSensitive to the individual Interviewer/observer bias (interpreting what they expect rather than interpreting objectively)Very detailed and in-depth information is gained. Nothing is overlooked.Lack of scientific validity no cause/effect.High ecological validity.Cant generalise the resultsCan be used to investigate rare phenomenon that would be unethical to use experiment for.Impossible to replicate lacks reliability.Reports of past events cannot be validated self report techniques are not reliableRevisionScholar

Work through the Unit 2 section of Scholar, adding to your notes and completing the interactive activities. Particularly the data analysis nominal, ordinal and interval data section which we havent covered.

1,3,4,2