D6vsczyu1rky0.cloudfront.net



ReliabilityReliability: In science, researchers should be able to measure something again and again & get the same/similar results. If they are consistent (get the same result) then they are reliable.Type of ReliabilityHow to assess itHow to improve itInternal Reliability: the extent to which a measure is consistent within itselfThe split-half method is one way of testing the internal reliability of self-report methods e.g. questionnairesThe internal reliability of observations is tested by assessing inter-rater reliability.Questions are removed and the split-half method is used again to see if internal reliability has improvedObservers should be trained and allowed to practice the use of coding systems/behaviour checklistsExternal Reliability: the extent to which a measure varies from one use to anotherYou can assess the external reliability of experimental and non-experiment methods by doing test-re-test. You can improve the internal & external reliability of an experiment by doing: A pilot study (to check the method of measurement works properly & that the ps can use them properly)Definitions:Split-half - You split the ps into two or more groups (depending on the different conditions being measured) & test them between the different conditions one at a time. If the tests show similar results from each test group between the different conditions it can be argued the test has internal reliability. If the results are vastly different between conditions then the study lacks internal reliabilityInter-Rater Reliability – This refers to the degree by which two (or more) observers agree on the data that is produced. It is measured by calculating the correlation between the two results. A general rule is that if there is more than 80% agreement, the data is said to have inter-rater reliabilityTest-Re-test - This see’s the same participant(s) being tested over a period of time on two or more separate occasions using the exact same procedures as previous. If the results are the same or closely similar you can argue the experiment has external reliability. Repeating the study/experiment between cultures and finding similar results would mean the study also has external reliability.Example Exam QuestionValidityValidity: Is the experiment measuring what it is supposed to measure? If it is measuring what it is supposed to measure, then it is validInternal Validity (experiments): Did the test used measure what it was supposed to measure within the experimental setting?E.g. Did the IV produce the change in the DV or was the change in the DV a result of a confounding variable? If a study is well-designed and extraneous and confounding variables are controlled then internal validity should be high. So, every experiment has an IV and a DV. Ideally, all other variables will be controlled so that they can’t vary systematically alongside the IV, affect the DV and bias the results. Any variable that does this is called a confounding variable. The presence of confounding variables affects the validity of the experiment. Controlling Confounding Variables: For example: In an experiment some ps learn a word list in noise, others learn the list without noise and the researcher records the no. of words recalled. There are lots of possible confounding variables EG, some ps may be tested in the morning and others in the afternoon, the researcher may give better instructions to the silence group than the noise group etc. Say it worked out that all the ps in the silence condition were learning in the morning while all the ps in the noise condition were learning in the afternoon. People are generally more alert in the morning and so should perform better due to the time of day. However, the researcher would claim the change in the DV was caused by the lack of noise and so time of day has become a confounding variable. The experimenter’s conclusions would therefore be invalid. Confounding variables reduce the internal validity of a study and so must be controlledExtraneous variables: These are ‘extra’ variables in the study. They do not vary systematically with the IV and so do not bias the results. However, they may affect the DV and so they are a nuisance as they get in the way e.g. if some ps get distracted during the study. Extraneous variables make it more difficult to detect an effect and so ideally they should be controlled. TABLE: Variables which could be confounding variables which influence the internal validity of the study & how to deal with themConfounding/ Extraneous VariableExamplesSolutionsParticipant variablesIQ, age, gender, motivation, ability, experience, personality etc.EG if, on average, ps in one group are more intelligent, or older, than ps in the other group, the intelligence/age difference may be the cause of the results rather than the IV. Incorrect conclusions may then be drawn. - Use a matched pairs design - Assign ps to groups randomly so that the chance of individual differences is lower. Situational variables Temperature, lighting, time of day, noise levels. Order effects (if repeated measures). -Keep all aspects of the situation the same for all ps- Counterbalancing to cancel order effects. Experimenter variablesResearcher biasA researcher’s expectations and behaviour can affect the outcome of a study. The bias can be direct - the researcher may behave in such a way that he/she gives cues to ps, thereby encouraging them to behave/respond in a certain way and therefore confirming the researcher’s expectations/hypothesis. So, a researcher’s cues can act like a confounding variable as a researcher may spend more time with some ps than others or be more positive with some ps than others. Also, a researcher may lead a p towards giving a certain answer. Researcher bias can also be indirect - a researcher may even create bias by designing an experiment in a particular way in the first place, or by operationalising variables in a certain way to fit what they want-Experimenter should act in a neutral manner - Experimenter should give standardised instructions so that all ps receive clear, equal instructions. - Double blind procedure where the investigator who meets the ps is unaware of the hypothesis and/or which group the p is in. -Pilot studyParticipant effects Social desirability biasIf ps know they are taking part in a study, they may alter their behaviour/response in order to portray themselves in a more positive light. Demand characteristics These are features of an experiment that provide cues as to the aim/hypothesis of the experiment. If ps guess the aim of the experiment, they may behave artificially, either to please the experimenter or to sabotage the experiment (the ‘screw you’ effect). Demand characteristics may become a confounding variable. Use a field experiment so ps are unaware of the research. - Stress confidentiality/ anonymity for the ps & importance of truthful answers - Single blind procedure, where ps don’t know the true aim of the study/are unaware of which group they’re in. (double blind would also be an option)-Experimental realism-Pilot studyStimuli variablesWord length, task difficulty etc-Independent measures; use same stimuli for all ps. - Repeated measures; match stimuli so they’re equivalent. Task: Read the following research study, then fill in the boxes:The investigation aimed to find out if memory is better when information is written in red than when it is written in green. The researchers took 2 groups and gave them the task of learning some pairs of words eg dog-table. Group 1’s words were written in red, while group 2’s words were exactly the same but written in green. At recall, ps were given the cue word (dog) and required to give the associated word (table). The number correct was recorded. 217170042545DV - 00DV - -57150042545IV - 00IV - 114300076835001485900-3810003086100144780What would the researchers have to do to control the study?00What would the researchers have to do to control the study?2667001905Possible Confounding Variables00Possible Confounding VariablesExternal validity (experiments): is the extent to which research findings can be generalised to other: Settings (ecological validity). For example the ecological validity of findings from an experiment in a laboratory tends to be lower than that of experiments in the field (though that is not always the case). Other groups of people (population validity) Different points in time (historical validity). Mundane realism also effects the external validity of experiments. Mundane realism concerns how much the experiment is like the real world. If it lacks mundane realism, it doesn’t mirror the real world & therefore lacks validity. For example the Loftus & Palmer study involved watching clips of car accidents. This is not like an everyday experience therefore the experiment may not be useful in understanding behaviour in the real world.Ways of dealing with issues of validity:Using a real life setting – in the field (more ecologically valid)EcologicalUsing natural tasks/stimuli – improved mundane realism & ecological validityUsing a random sample (from different areas / social groups) – improved population validitypopulation Note – although internal and external validity are different things they are linked. Results cannot be generalised outside of the experimental setting if internal validity is low, as they may not be true. Exam Style Questions:Explain why it is important to control extraneous variable in a study [2]Distinguish between extraneous variables and confounding variables [3]Explain what is meant by demand characteristics [2]Explain what is meant by researcher bias. Use an example in your answer [2] ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download