Lecture notes, lectures 1-18 - Lecturer was dr. julie mattiske PDF

Preview

Summary

Lecturer was Dr. Julie Mattiske ...

Description

• Research ◦ Gives us knowledge about human behaviour, what works and what doesn't ◦ Increases our knowledge/understanding of something ◦ Good answers come from good methodology ‣ Bad answers come from bad methodology • Statistics & data analyses ◦ What do the results mean? Do they matter? • When researchers/practitioners make claims of fact ◦ We she evaluate the evidence ◦ Trust ourselves to evaluate, not just believe what 'experts' say ‣ Researchers are people, people make mistakes • Psychology research ◦ Allows us to test and justify claims ◦ Very important to say what is right and what is wrong, and to not expand on the truth ◦ Report evidential facts ◦ Reputable research (research with good reputation) ‣ Published in peer-reviewed journals (researcher has a research question they are interested in, designs a study, measures, makes decisions about methodology, collects data, analyses results, writes up a manuscript intro, method, results, discussion, sends off to a journal) ‣ Researcher, based on the quality of the study, send it to the best journal they think that will accept it, editor assesses it and sends it to reviewers who have knowledge in the area - the reviewers assess whether it is publishable ‣ This is why peer reviewed studies are the best as they have been through the above process • Thought field therapy ◦ An illustration of claims about causes of people's problems & about therapy ◦ Invented by Roger Callahan ◦ Involves tapping on energy points on the body

Causality & Internal Validity • Internal validity ◦ Closely related to cause and effect ◦ Definition from conducting research in psychology text: ‣ The degree to which a research finding provides accurate information about causality. The degree to which changes in the IV in a particular study really do influence the dependent variable in the way suggested by the results of the study ‣ When a study is high in internal validity, we can confidently conclude that variations in the IV caused any observed changes in the DV ‣ Crucial for testing: high internal validity = changes in IV caused changes in DV ◦ Why do we care about cause & effect? ‣ If you know what a cause is, a treatment can be identified ‣ To identify why people do things instead of what they do • Experiments ◦ Can demonstrate cause and effect ◦ Participants randomly allocated to conditions/groups ◦ Experiments allow cause & effect conclusions ◦ Good experimental design holds everything constant except for the IV • Non-experimental studies ◦ Correlations & comparisons of pre-existing groups ‣ People who are already unemployed compared with people who already have a job ◦ Sometimes we can't always randomly allocate participants (can't do experiments) ◦ Problems with interpreting correlations ‣ Cannot rule out alternative explanations ‣ Therefore cannot determine cause & effect ‣ Cant demonstrate causality ‣ Thus focuses on associations and relationships ◦ Example of non-experimental study and problem with causality ‣ Having close friends who are positive affects someones wellbeing, by providing support when needed • Social support --> wellbeing ‣ However, being depressed may result in less interest for friends, which affects someones social life • Wellbeing --> social support ‣ Or, an alternative explanation could also be responsible for the association between social support/wellbeing • A and B are not causally related, rather they are associated/ influenced by another variable (C) • Going through a rough patch might influence one to see everything

•

•

•

•

•

negatively (negatively viewing friends and their general psychological wellbeing) Essential conditions for establishing causation/causality ◦ Covariation: 2 events (cause and outcome) should be related to each other ‣ Higher social support should be related to greater wellbeing ◦ Temporal order: cause must precede (come before) the effect ‣ Correlational studies do not tell you this ◦ Elimination of plausible alternative explanations for the covariation ◦ Well designed studies can achieve all the above The logic of experimentation ◦ Presenting two conditions with everything the same except for one component (IV) ◦ To conclude A = B ◦ Everything is identical, only difference is the outcome (whether people got B or not) ◦ Good experiments hold everything constant except for the IV, however this is not always possible Well-designed experiments ◦ Can rule out alternative explanations ‣ Good internal validity Internal validity ◦ The extent to which we can say any effects on the DV (wellbeing) were caused by the IV (social support), as opposed to something else (alternative explanations) Threats to internal validity ◦ Threat = potential alternative explanations that cannot be ruled out ◦ Threats to internal validity are controlled by the set up of the experiment prior to collecting any data ◦ Threats are controlled by ensuring they apply similarly to different conditions ◦ Types of threats ‣ History • Collecting data over time • Something might happen outside the study during this time, that has potential to impact results • Affect relevant group, not an individual • For example, doing a study of the psychological impacts resulting from smoking, and a famous celebrity dies from lung cancer and starts a campaign saying 'don't smoke' • Prepare for threats by thinking about occurrences of possible events in regards to your experiment • How do you control for history threats? ◦ Have a control condition ◦ History effects will apply to all, as all are measured at the same time and in the same environment

◦ Measure at baseline, measure at end Maturation ‣ • Naturally occurring changes in people, over time ◦ Children get older, their vocabulary increases, cognitively enhanced • Example - money management program for year 12 students ◦ Measure at baseline, measure 12 months after ◦ In the meantime, students could have gotten a job, could have attended uni, live independently ◦ Control for this with a condition that doesn't (get a job, attend uni, live independently) • Example - common cold ◦ One has a common cold, kerosene is thought to treat the common cold ◦ Have a sniff of kerosene and the common cold diminishes ◦ Does kerosene cure the common cold? ‣ No, because our immune system naturally cures the common cold ◦ How do you control for this? ‣ Have a control condition who does not sniff kerosene (and a condition who sniffs kerosene) ‣ Must be measured over the same time (in both groups/ conditions) ‣ Regression towards the mean • Say something brief about this in assignment ◦ Buttner et al.'s control condition controlled for this • Based on observation • If something is measured twice in the same person, it is common for the 2 scores to differ. Scores that are very different from the mean on a first measurement occasion can be nearer to the mean when the measure is taken a ◦ This is because of ‣ Errors of measurement - we cannot exactly measure emotions (we cannot put a microscope into someones head and determine how excited, depressed, unhappy someone is) ‣ Chance events - little things can happen to a particular individual that can affect their score on a certain occasion (might be in a bad mood/good mood could affect performance on a cognitive task) ‣ Extreme scorers (high scores) may move closer to the mean, rather than further away (if their score was initially 80 with the mean at 47, they will move closer to the mean in the second measurement with a score of 60) • To control for this: take repeated measures in both a treated an

control sample, regression to the mean should apply similarly Selection ‣ • Anything that can result in non-equivalent groups in different conditions, reflects the method of condition assignment • Example - Volunteers say 'great! yoga! I'd be more than happy to participate', non-volunteers say 'yuk, yoga' ◦ 2 months after the yoga intervention, yoga group is happier/ healthier than non-yoga group ◦ What is wrong with this? ‣ Yoga group may initially be healthier/happier - already attracted to yoga/health/relaxation ‣ However, those who put their hand up for yoga may also be people who are motivated to finally do something about their health/stress levels ‣ Non-random allocation involves a bias towards yoga (nonequivalent groups) • Selection threat involves pre-existing differences between conditions • Preventing selection threat ◦ Randomly allocate people who are prepared to do yoga - spread out the people who are committed to doing something about their health & those who like yoga ◦ To control for this, randomly allocate participants to different conditions (don't confuse with randomly selecting participants) ‣ Confounds • You & I are running a study example ◦ Can you conclude that cognitive therapy is more effective than behaviour therapy? ‣ No, although the results of cognitive therapy are better than behaviour therapy, the difference in results could be attributed to a confound - the CT psychologist is better than the BT psychologist ◦ The confound is = different therapist for different therapies ‣ As a confound is something that differs between conditions • Has to be related to the IV ‣ It is not an uncontrolled variable that applies to different individuals • For the assignment: we use the term confound for additional problems (of internal validity) that aren't covered by one of the other labels ◦ Some confounds occur because of a weakness in the study design

• Internal validity ◦ Threats to internal validity ‣ How they can be controlled • Potential threats are controlled if they apply similarly in control condition/experimental condition • If this is true, the only difference between conditions is the IV • A good study design must have the IV as the only difference between conditions, as if there are other differences besides the IV it will be hard to determine whether the IV affects the results ‣ Ravens progressive matrices • Measure of visuo-spatio reasoning • People work out what is next in the pattern • Example: test the affects of the tesla shield with ravens progressive matrices task ◦ Has the tesla shield made her smarter? ◦ No, (refer to slide You test the Telsa Shield on me) because she has had practice as this is the second time she performs on the Ravens Progressive Matrices task (baseline & post-treatment measures) ◦ This refers to testing effects/practice effects ‣ Testing effects/practice effects • People do better on a performance measure because they have already had an attempt at it • Changes that occur in a measure/score when it is tested on two or more occasions (they have learnt from the first test), because of the earlier measurement • How do you control for this? ◦ Have a control group (no tesla shield): perform the task at baseline and post • Don't only apply to ability measures - also apply to personality measures • When people fill out measures it prompts them to think about and consider what is being measured ◦ For example, an intervention is administered on parents who spend a lot of time at work and less time with their children. The intervention is administered at baseline and post, however, in between the administrations (a month apart) the parents actually consider how they spend minimal time with their children. After considering this they then begin to spend less time with work and more time with their children. Thus, this is not suffice, a control group is needed for this example • It is important to control for when measures are taken twice ◦ So the results are true, and not because of another reason ‣ Instrumentation

• Refers to measures requiring skills or discretion (judgement) or effort • Changes in the way a measure is applied or interpreted over time • The data collected between baseline and post intervention may differ if the person administering the intervention is underprepared at baseline, but is well prepared at post ◦ This is because there will be changes in the way a measure is interpreted over time • Can even apply to self-monitoring participants where they are asked to keep a daily diary ◦ Quit smoking program = researcher asks participants to fill out their diary every time they have a smoke (time, situation, mood, alcohol) ◦ At the start of the program they fill it out every time they have a smoke ◦ However, as the novelty wears off they might start to forget taking the diary with them and fill it out when they are at home at night where they have to sit down and really think about the details they have to fill in ‣ This is an example of change in effort towards the measure over time ‣ Demand effects • Participants produce results based on knowing or guessing what the researcher is looking for • Can produce bias in the way measures are completed • Responding in a way that is consistent with the researcher's expectations • Participants ideas about what researcher expects ◦ Threats to internal validity that apply to intervention studies (studies on treatment) ‣ Intervention studies can involve a vast range of treatments ‣ Non-specific treatment effects • Improvement following therapy ◦ Responses are different at post-therapy measurement compared to baseline ‣ Is this necessarily because of the treatment? - this is the real question ◦ There are many other things that come with a face-to-face psychological therapy that the client receives (rather than from the specific treatment itself) ‣ Non-specific aspects of therapy (warmth, acceptance, support, reassurance, education, homework exercises) ◦ Could these aspects alone lead to improvements in scores even if the therapy isn't useful? ‣ Absolutely - in other words, this says our therapy is

absolutely useless but everybody at the end of the treatment is feeling better • This could be due to non-specific aspects of therapy ‣ For example, getting educated by the therapy, receiving warmth/support, talking to someone who listens, receiving acceptance/reassurance ‣ All of these factors have nothing to do with the therapy itself (non-specific aspects of therapy), however, these aspects may be the reason of improvements in scores (may not necessarily be the therapy that is the reason of improvements in scores) ◦ Improvements that occur owing to non-specific aspects of therapy ‣ Placebo effects • If a therapy is credible and the participant trusts the researcher, the expectation of the participant might be that they will improve ◦ In other words, people feel better just because they know they're receiving treatment, not because of the effectiveness of the treatment ◦ Placebo effects result from the participants own expectations about what will happen • The expectation could lead to improvements in scores regarding a therapy ◦ Having positive expectations could lead to improved scores on a therapy (when people expect to get better, they therefore seem as if they have) • Thus, placebo effects could have contributed to the 'improvement' • The graph of 'Placebo effects in depression' ◦ Demonstrates drugs used on people who clinically suffer from depression (including a placebo) ◦ The placebo involves a fake drug (empty capsule) that looks the same as the active drug ◦ Over a period of 6 weeks both the active drug and the placebo show improvements in the severity of depression ◦ The placebo has resulted in improving to a mild case of depression, which is a good outcome for what is an empty capsule ‣ How do we control for these threats to internal validity? • Have a control & make sure that these threats apply similarly to both conditions • Expectations/hope aroused, attention, warmth from researcher, delivery of information/treatment and any other non-specific aspects of therapy should all be the same for all groups ◦ Placebo treatments should also be similar to active treatments in credibility

• Want to test if the Telsa Shield enhances physical and mental energy as advertised ◦ Have a control condition ◦ And could find something that looks similar to the Telsa Shield (treatment condition) ◦ By doing this you would control for placebo effects/demand effects/non-specific treatment effects • You wouldn't have a control condition that does absolutely nothing ◦ 'Make sure these threats apply similarly to both conditions' (control/experimental) ◦ As you unveil the true effects/results of whatever is tested • What about controlling threats to psychological interventions? ◦ Is it possible to have a placebo? Use a psychological placebo ◦ Goldstein et al tested whether EMDR was beneficial for panic disorder with agoraphobia ‣ Active treatment: received EMDR ‣ Attention placebo: told participants about ART, combined things that are useless for panic disorder but tried to make it seem credible ‣ Waiting list control: don't get any treatment until study is over ◦ EMDR & ART groups ‣ Placebo effects = same, found that EMDR and ART were credible - both groups had the same expectations (that they will improve) ‣ Non-specific treatment effects = same, amount of contact/ attention/warmth was same for both of the groups ‣ Demand effects = both groups had contact with the researcher and therefore could guess what the researcher was looking for ◦ Waiting list control - did it control for? ‣ Placebo effects = no, not the same, they received no therapy ‣ Non-specific treatment effects = no, not the same, didn't have contact with the researcher ‣ Demand effects = no, not the same, no contact with the researcher therefore didn't know what was being looked for ◦ Outcome = active treatment (EMDR) & attention placebo (ART) were no different ‣ Thus, EMDR treatment was not useful ‣ This is why active placebo controls are used - so we can tell between therapy effects and expectations and the mechanics of delivering therapy ‣ Effect on self-reported steadiness • Participants were shocked with a minute electrical wave and were told the wave will help anxiety/fatigue/concentration

• Some participants were told they would feel steadier or unsteadier when they stand in a still position • A vast majority of those who were told they would feel steadier said they did feel steadier • And a majority of those told they'd be unsteadier reported they were unsteadier than normal ◦ Both of the above could be result of either placebo effects (participants own expectations) or demand effects (participants ideas of what the researcher expects) ◦ Threats to internal validity (general - not applying to interventions) ‣ Attrition (mortality) • If studies run over time and take measures on the same people attrition is common/ordinary • If you collect data over a period of time, people will drop out of the study • Drop outs are usually not a problem, but it depends on the pattern/ causes of drop outs (drop outs are people who leave the study before all intended data is collected) • E.g., people might drop out because the intervention/treatment is not working ◦ So, you are stuck with data from people with whom the intervention worked ◦ There will therefore be a bias in the results ◦ Or, people may drop out because the intervention has worked, and you will then have a bias in the results in the other direction • Attrition is generally reported ‣ Experimenter bias • Administering instructions differently in different conditions • Could bias results • Where instructions are not uniform throughout the whole experiment by result of bias from the experimenter • Discretion or judgement - bias/scoring more favourably for one condition more than another • Overall, how do we control for potential threats ◦ Ensuring they apply similarly in all conditions ◦ That one condition is not different to another except for the IV ◦ Do all treatments need a placebo? ‣ No ‣ When measuring the effectiveness of a treatment - Yes ‣ When you want to know what treatment is better - No ◦ Do all studies need a control condition ‣ No ‣ What are relative effects of treatment A & B? - No ‣ Which memory strategy works best out of 3 strategies? - No

‣ Correlation studies - control not relevant ◦ Expectations/hope aroused, attention, warmth from researcher, delivery of information/treatment and any other non-specific aspects of therapy should all be the same for all groups ‣ Placebo treatments should also be similar to active treatments in credibility

Design and Co...