RESERCH METHODS Psychology IB1 09/2017-10/2017 Psychology is a scientific study of mental processes and behaviour. This means theories and studies: 1. Should be supported by empirical evidences and based on them. 2. Should be falsifiable > can be proven wrong! 3. There is a need for history of independent attempts to replicate and support theory > one isn’t enough! Wrongly controlled or badly designed attempts can produce artefacts > result associated with the effect of unseen factors. Theory > it is an explanation for psychological phenomena. Theory also makes prediction about future behaviour and is used to summarize, organize and explain observations. Most theories are built on concepts > hypothetical constructs, carefully defined so that they can be tested. Basis for good theory includes teacup: 1. 2. 3. 4. 5. 6. Testability > are we able to test it and state if its “true” or not? Empirical support > replicated! Single study isn’t a good support, there is a need for various evidence. Applications > high heuristic value > theory can be applied to explain various behaviour Clearly defined variables and concept > can be reliably measured in scientific way Unbiased > various biases can occur, we need to get rid of them to make a good theory Predictive > theory should not only explain behaviour, but also predicts trends within a population To support psychology uses empirical data (evidences collected through scientific methods). Anecdotal data (casual, collected in informal manner, relying on personal experience) cannot be used in scientific research! Behaviour > everything that can be registered by independent observer – actions, gestures, verbal expression. Mental processes > things that stay behind the scene > it includes e.g. attention, perception, memory, thinking. Though they cannot be observed directly, they can have effect indirect influence of them on behaviour. In psychology, variety of research methods is used, in order to support or refute theories and gain “evidences”. Key point > psychology does not and cannot prove anything! It can either support of refute proposed theory. To achieve that, psychologist use variety of research method. Key points in preparing for study and research: - Aim - Method – Procedure - Results – Conclusion - Evaluation Sample > group of individuals taking part in research study > sampling is the process of recruiting them Generalizability > extent to which results can be applied beyond the sample and setting used in study Credibility > degree to which the results of study can be trusted to reflect reality > closely linked to bias Bias in research > a lot of traps for researchers! If there was a bias in research, results don’t reflect reality There is a need for accurately designed procedure > proper type of method needs to be chosen. Include: Location of study > lab based or field study Retrospective or prospective > researcher can either ask about past of subjects (retro) and look for correlations or measure the variable at beginning of study and watch the effect of variable over time RESERCH METHODS 1 Longitundal or cross-sectional research > study may involved repeated observations of the same variable over long periods or researched can analyse data collected from a population at specific point in time. Ideally, theory should be supported by both! Type of data which we want to obtain > qualitative or quantitative. Both have disadvantages and advantages and provide us with different types of data, different potential and different costs In preparing the study, there is a need of getting rid of all additional cofactors that may affect the results. Main things to do > state the aim and decide what we want to obtain! QUANTITATIVE_____________________ Quantitative methods operates with variables > all things that can be objectively registered and quantified. All variables should be controlled during experiment, so the results are as accurate as possible > need for Construct > any theoretically defined variable (i.e. aggression) > to define construct, we need to clearly delineate it from other similar constructs > basing our explanation and definitions on theories. As a rule > constructs cannot be directly observed > we “construct” them based on theory. Operationalization > to enable research, we need to operationalize constructs > express it in terms of observable behaviour (i.e. the number of insults per hour) > operationalization can be creative! Sometimes, it is hard to operationalize constructs > how can we objectively measure e.g. love, belief in God? Main types of quantitative research include descriptive, correlational and experimental studies. 1. EXPERIMENTAL STUDIES > often simply changing IV and DV with other potentially important variables controlled > we measure how manipulated IV changes DV. Only type measuring cause-and-effect! Independent variable (IV) > variable that is manipulated during the experiment > can have effect on DV Dependent variable (DV) > variable which is measured by experimenter (can be affected by independent V.) Extraneous (confounding) variables > other variables, which aren’t IV but could affect results or affect DV As everywhere, in a experiment, there is a need for good sample. It should be representative of target population > population to which we want to generalize our results. Choice of sample must be justified! Types of sampling in experiment: a) Random sampling – perfect approach to make the sample representative – every member of target population has equally big chance of being chosen to experiment. There is still a need for sufficient sample size. It would be perfect, but we cannot always ensure that each member of population gets equal chance to enter (e.g. we can’t call someone from China to come - we can either belief that cross cultural differences aren’t essential here, or narrow target population) b) Stratified sampling – we decide which characteristic should sample reflect and then study the distribution of them in population (e.g. age distribution) > the we recruit the participant that keeps the same proportion as the one observed in target population. c) Opportunity (convenience) sampling – recruiting participant that are more easily available. d) Self-selected sample – recruiting volunteers > quick and easy, but often low representativeness Designing the experiment > good procedure must be chosen. It can affect validity! Independent sample design > involves random allocation of participants to conditions and comparison within groups. We manipulate with variables so that conditions are identical (except of IV) and compare the dependent variable in two groups. Potential confounding variables cancel each other out. But if groups RESERCH METHODS 2 weren’t the same – we are comparing oranges to apples. Though, is size of sample is sufficient and allocation is completely random, chances that group are equivalent are higher. Of course, there can be more than on IV! Matched pairs design > instead of randomly allocating to groups, researcher use matching to form them. E.g. we prior to experiment assess the level of variable that can confound (e.g. earlier abilities) and allocate participant equally. This controlled variable is called matching variable. This is preferred for small sample! Repeated measured design > it is used when the goal of researcher is rather to compare conditions than participants. The same group of participants in exposed to two (or more) conditions and results are compared. Problem with this type > they are vulnerable to order effect > results can be different depending on which condition is first (because of participant fatigue and practice – participants improve, become more confident). Can be overcome by counterbalancing > usage of other group, where conditions are reversed. This type of design overcomes the influence of participants variability > differences between groups. VALIDITY OF EXPERIMENTS Construct validity > moving back from operationalization to construct can be tricky > this leap between them must be justified. Remember, that e.g. swearing does not always mean high level of aggression. Internal validity > methodological quality of the experiment > were the confounding variables controlled, are we certain that it was the IV that caused the change in DV, etc. Threats to internal validity includes biases External validity > generalizability of findings. There is an inverse relationship between internal and external validity – need to decide to what extent study should the study be artificial, controlled, standardized - the more it is, the lower the external validity. We can distinguish two types: Ecological validity – extent to which finding can be generalised to conditions other than experimental Population validity – extent to which results of sample can be generalized to populations BIAS IN EXPERIMENTS 1. Selection – this occurs when groups weren’t equivalent at the beginning 2. History - refers to outside events that happen to participants during experiment, which can influence DV 3. Maturation – during experiment, participants go though natural processes – fatigue, growth etc. 4. Testing effect - the first measurements of the DV may affect the second measurements –it can happen e.g. when participants need to do the same things – on second attempt, they are already familiar with it. 5. Instrumentation – refers to slight change of instruments during experiment. Relevant, as in most cases, the observer is the instrument of measurements > he collects data, observes, etc. > sometimes tools! 6. Regression of the mean – if initial score of DV is extremely low/high, probably it’ll become more average 7. Experimental mortality – refers to participants dropping out during the study – dropouts aren’t random! 8. Demand characteristics – when participant understands purpose of study and tries to fit their behaviour into experimenter’s expectations. We can distinguish good-participant role (trying to discern hypothesis and confirm it) and negative-participant role (discern and destroy credibility) – also known as screw-you effect. 9. Experimenter bias refers to researcher unintentionally exerting influence on results. To avoid it, we can use double-blind design, where neither participant nor researcher know which group is control/experimental TRUE LABORATORY EXPERIMENTS RESERCH METHODS 3 In lab experiments, the independent variable is manipulated by researcher - so causation can be infer. Lab provides us with better control of confounding variables, providing higher internal validity. QUASI EXPERIMENTS Allocation of groups in this case isn’t random – instead, pre-existing differences between groups are used. Major limitation of quasi experiment > it is impossible to determine cause-effect relationship, as we aren’t sure of equivalence of comparison groups at the start of study (in all variables). Can be in lab! E.g. researching anxious and non-anxious group - we can’t be sure if anxiety is the only difference FIELD EXPERIMENTS Similar to lab, but conducted in different settings. This real-life environment provides high ecological validity. Though, as many extraneous variables cannot be controlled, the internal validity is lower. NATURAL EXPERIMENTS Just like field experiments, these are conducted in natural environment. Though, researcher cannot control the IV – it is manipulated by nature. All natural experiments are quasi experiments! 2. CORRELATIONAL STUDIES > no variable is manipulated > we cannot infer causation. Here, two or more variables are simply measured and their relationship is quantified mathematically. Correlation – a measure of linear relationship between two variables. Can be either positive (one rises when other increases) or negative (one increases while the other decreases). The steeper line – stronger correlation. Effect size is the absolute value of the correlation coefficient. [<0.10 – negligible; 0.3-0.49 - medium; 0.5> - large] Statistical significance shows the likelihood that correlation has been obtained by chance. [P < 0.05] LIMITATIONS OF CORRELATIONAL STUDIES Correlations cannot be interpreted in terms of causation > we cannot determine what causes what Curvilinear correlations – sometimes, variables are linked non-linearly. E.g. there is a positive correlation, but only up to some point – then, there is negative - we cannot simplify it into straight line The third variable problem – always a possibility for presence od 3rd variable, which explains others Spurious correlations – if researcher calculate multiple (e.g. 100) correlations, the statistical significance (5%) can be not enough > this five of our hundred results could be artefacts. Sampling methods used in correlational studies are the same as in experiments. Target population is found depending on aims. The generalizability is directly linked with sampling, and depends on representativeness. BIAS IN CORRELATIONAL STUDIES In this kind of research, bias can occur on various levels – in measurement of variables, interpretation of results, in calculating correlations. QUALITITATIVE____________________ Qualitative research methods focus on in-depth study of a particular phenomenon. In-depth means that research goes beyond what can be objectively measured and quantified. Qualitative research includes of interviews, observations, focus groups, case study, and content analysis. Interpretation of such data involves degree of subjectivity, but achieved by it analysis is deeper than the one obtained from quantitative research. CREDIBILITY IN QUALITATIVE RESEARCH RESERCH METHODS 4 Credibility is an equivalent of internal validity in experimental methods. Similarly, here, credibility in related to question “To what extent do findings reflect reality?”. To denote it, we also use term trustworthiness. To ensure that the findings are as true as possible, few types of measures can be taken: 1. Triangulation – refers to combination of different approaches to collecting and interpreting data. Several types of triangulation can be used Method triangulation – use of different methods in combination can compensate for their individual limitations and reinforce their strengths > if same results are obtained using various methods (e.g. interviews and observations) the credibility increases. Data triangulation – using data from variety of sources. E.g. studying documented biographical data Researcher triangulation – combining observations of different researchers > two people see the same Theory triangulation - using various perspectives or theories to interpret data 2. Establishing a rapport – researcher must ensure that participant are being honest > he should reminds subjects about voluntary participation and right to withdraw – so that responses are from ones that are willing to contribute. It should be clear for participants that there is no such a thing as good/wrong answer. 3. Iterative questioning – spotting ambiguous answers by rephrasing the questions 4. Reflexivity – researcher should reflect – did his biases interfere the interpretation/observation? We can distinguish epistemological reflexivity (strengths and limitations od used method > i.e. participant knew they were being observed) and personal reflexivity (linked to personal beliefs/expectations of researcher) 5. Credibility checks – checking data accuracy by asking participant to read and confirm collected notes 6. Thick descriptions in data - explaining not only behaviour but also context of it > so that it can be understood holistically > using sufficient details for descriptions etc. BIAS IN QUALITATIVE RESEARCH Bias in always present, as here, the researcher is the tool used for research. Some types of bias are inevitable. We can distinguish participants and researchers bias > both should be avoided. Participants bias includes: 1. Acquiescence bias – tendency to give positive answers, whatever the question is. 2. Social desirability bias – participant’s tendency to behave in such a ways that they think will make them liked/accepted. Can be unintentional and intentional, in order to e.g. try to help the researcher 3. Dominant respondent bias – in groups > one participant influence or intimidate others responses 4. Sensitivity bias – tendency of participants to avoid and distort responses to sensitive questions. Researcher bias includes: 1. Confirmation bias – when researcher has prior beliefs and unintentionally tries to confirm them. It can be: order of questions, selectivity attention, or disregarding data that don’t match with the belief 2. Leading question bias - participant is incline to answer in certain way because of the wording used 3. Questions order bias occurs when response to one question influence response to the following ones 4. Sampling bias occurs when the sample isn’t adequate to the aims of research 5. Biased recording – when findings of study are not equally represented in research rapport SAMPLING IN QUALITATIVE RESEARCH In qualitative research, sampling is non-probabilistic > in opposite to quantitative one. We can distinguish: RESERCH METHODS 5 a) Quota sampling – prior to study, researcher decides how many people should he include and what characteristics should they poses. The, he recruits participant until the quotas are met. b) Purposive sampling – same as quota, but proportions and sample size aren’t defined. c) Theoretical sampling – recruiting, until data saturation (no new data can be obtained) is reached. d) Snowball sampling – small group of people is invited and asked to invite others with similar traits e) Convenience sampling – using sample which is easily available and accessible GENERALIZABILITY IN QUALITATIVE RESEARCH 1. Sample-to-population generalization – hard to obtain, because of non-probabilistic sampling 2. Theoretical generalization – made from particular observation to broader theory. Similar to the leap made in quantitative methods for construct validity – from direct observation to unobservable construct 3. Case to case generalization (transferability) – made to a different type of group in different setting. TYPES OF QUALITATIVE RESEARCH METHODS 1. OBSERVATION – the focus of this type of research is to find how people interact, and interpret each other’s behaviour. Only in observation real behaviour can be observed > in other, such data is often artificial. Used also when researcher believes that data can’t be articulated in other way (e.g. asking about behaviour during fire won’t give reliable answers – observation is used). We gain almost first-hand experience; researcher is deeply immersed into studying phenomena, sometimes becoming part of it (threat to objectivity). Biggest problem with observations: reactivity (participants behave differently just because they are being observed) LABORATORY VS NATURIALISTIC Sometimes, naturalistic observations are the only choice (e.g. studying cult’s behaviour or aggression – it would be unethical to artificially create it). Drawback – it can be hard consuming – need to wait OVERS VS COVERT When participants are aware of being observed (overt) the study is ethical (consent can be given). Though, results can be interfered by biases and [un]intentional changes in behaviour. In contrast, covert gives admission to groups otherwise inaccessible – but ethics are a disadvantage here. PARTICIPANT VS NON-PARTICIPANT Researcher can immerse himself into observed group, becoming a part of it. This type allows the researcher to gain first-hand experience, but subjectivity and ethics are threatened. STRUCTURED VS NON-STRUCTURED In structured observation, data is recorder systematically and in standardized way. 2. INTERVIEWS are popular types of qualitative research methods, because of few advantages: they can be the only way to get an insight into nature of subjective experience/interpretation; they can be used to understand the meanings participants attach to certain event/points of view; in depth-interviews can be used when topic is too sensitive for people to discuss it in group. Collected data are interview transcripts and notes Structured interviews include list of fixed questions, that need to be asked by researcher Semi-structured do not specify order or particular set of question > there is more a checklist Unstructured are mostly participant-driven and questions are determined by previous ones/answers Focus groups are special type of semi-structured interview, conducted simultaneously with a group of 6-10 people. Key factor – participant are encouraged to interact with each other > the interviewer is more a facilitator. It is easy and fast method, but anonymity and confidentiality are harder to preserve. RESERCH METHODS 6 Content analysis – recordings of transcript need to be analysed > need to avoid researchers subjectivity. This is done by inductive content analysis – looking for schemas, topics that reoccur. All findings must have their bases in transcripts and collected data, with maintained balance between observation and analysis. If theory emerges from data, it is grounded theory – “grown out” of empirical data, without prior beliefs 3. CASE STUDY is an in-depth investigation of an individual or a group. This can involve variety of methods (interviews, observation etc.) – anything, to deepen our understanding of a case. They are refereed as separate method because: object of study is unique – we want to gain understanding of this particular phenomena; sampling is not an issue; less focus on generalizability; various methods, often longitudinally. SUMMARY OF METHODS QUNTITATIVE QUALITATIVE Aim Nomothetic approach > derive universally applicable laws Idiographic approach > in-depth understanding of a particular case Data Numbers Texts Focus Behavioural manifestations (operationalization) Human experiences, interpretations, meanings Objectivity More objective More subjective ETHICS_______________________________ As psychology deals with living beings – this is what distinguished human sciences from natural ones. There QUANTITATIVE RESEARCH Sampling Generalizability Credibility Bias Experimental studies Correlational studies QUALITATIVE RESEARCH - Random - Stratified - Self-selected - Opportunity - Random - Stratified - Self-selected - Opportunity - Quota - Purposive - Theoretical - Snowball - Convenience Population validity Construct validity - Sample-to-population - Case-to-case - Theoretical No special term; validity and credibility used interchangeably Triangulation, reflexivity, establishing report, thick descriptions. credibility checks External validity Population validity Ecological validity Construct validity Internal validity Controlling confounding variables Threats to internal validity: selection, history, maturation, testing effect, instrumentation, regression to the mean, experimental mortality, demand characteristics, experimenter bias RESERCH METHODS Measurements: used techniques and tools Interpretation: curvilinear relationship, third variable, spurious correlation Participants bias: social desirability, sensitivity dominant respondent, acquiescence, Researcher bias: sampling, confirmation, leading question, question order 7 are ethical considerations in conducting the study as well as in reporting the results. In conducting the study: a) Informed consent – participant must be in study voluntary and fully understand the nature of it. b) Protection from harm – all time participant must be protected from physical and mental harm – includes possible long-term consequences c) Anonymity – data is anonymous, if no one can trace the results back to participants identity d) Confidentiality – if someone know who said that (e.g. researcher), data should be preserved in such way that only this one researcher knows who said what - confidential if he doesn’t share with those e) Withdrawal from participation – participant has the right to withdraw at any point f) Deception – sometimes true aims cannot be revealed to participants- they can change their behaviour g) Debriefing – after the study, participant must be informed about its nature, true aims and results. Cost-benefits analysis – should be conducted prior to the study. In all countries, there is an ethical committee that resolve ambiguities and approve research proposals. Ethical considerations of reporting the results include: 1. Data fabrication. 2. Plagiarism. 3. Publication credit. 4. Sharing research data for verification. 5. Handling of sensitive personal information. 6. Handling of information related to mental disorders. 7. Social implications of reporting scientific results RESERCH METHODS 8