Research Methods

Features of a science
-Contains testable theories
-It should be objective
-Methods should be empirical
-It should be valid and reliable
-It should be ethical
Some parts part of psychology are more scientific than others
-Biological = Uses empirical methods that are easily replicated and aren't affected by participant variables and is falsifiable = Very scientific.
-Behaviourist = Uses empirical methods looking at behaviour and is falsifiable but can be affected by participant variables = Quite scientific.
-Cognitive = Uses empirical methods and is falsifiable but it's hard to separate cognitive processes and can be affected by participant variables = Reasonably scientific.
-Social = Uses experimantal methods and observation getting qualitative data, making variables difficult to operationalise and control = Not very scientific.
-Psychodynamic  = Abstract concepts that can't be tested or falsified, producing unreliable data that can't be generalised or replicated = Not scientific. 

Validity: How well a test measures what it claims to.
-Internal is the extent to which results are caused by the controlled variables.
-External is the extent to which results can be generalised.
-Ecological is the extent to which results test real-life.

Validity can be assessed by looking at face (simply looking to see if it appears to measure things correctly), concurrent (comparing its results to an established test), or predictive (checking whether results will correlate with later results) validity. 

Reliability: How consistent and dependable a test is.
-Internal is when different parts of the test should give consistent results.
-External is when the test should produce consistent result regardless of when it is used.
-Inter-rater is when the test should give consistent results regardless of who uses it.

Reliability can be assessed using split-half method (splitting the test in half to see if results will correlate), test-retest (repeating a test later to see if it will correlate), or by having two researchers do the test (to check for bias).

Peer review

All scientific work has to be peer reviewed before publication. This is when other researchers check the quality of the work to ensure that only research that is up to a certain standard is published. Poor research won't be published, which makes published theories more valid and trustworthy. When research is published, scientists attempt to replicate it. If the same results are achieved then the theory is reliable as it shows that the results aren't affected by time or place.

Ways of improving validity and reliability

-Standardising research involves creating a specific procedure to be followed every time the test is carried out to ensure the test is done in exactly the same way.
-Operationalising variables involves clearly defining every variable. This can be giving the amount, age and gender of participants, or stating a specific test to be used for measuring the dependent variable to ensure the same materials and participants are used every time.
-Pilot studies are small scale trial runs of the research that allow researchers to check for problems and improve the test before the real thing. 

To ensure samples are representative of the population it must include various characteristics that are found in the group e.g males and females, young and old.

-Random sampling is when everyone in the group has an equal chance of being slected. It is the fairest type of sampling but doesn't guarantee a representative sample as some subgroups may be missed.

-Opportunity sampling takes advantage of whoever is available at the time. This is quick easy and cheap but is unlikely to be representative.

-Volunteer sampling advertises the test and invites people to volunteer. This often produces a large sample, however only a certain personality type are likely to volunteer.

Observation - Can be used when intervention would be inappropriate for example when animals are involved. Subject to experimenter bias as observed behaviours may be interpreted differently.
Lab experiment - Good control over extraneous variables and highly replicable. Lack of ecological validity.
Field experiment - High ecological validity and lack of demand characteristics. Deception used often seen as unethical.
Natural experiment - High ecological validity. Lack of control over extraneous variables.
Questionnaire - Quick and easy for obtaining large amounts of info. Self-report therefore risk of social desirability. Misunderstandings can't be clarified. Open questions subject to interpetation. 
Interview - Misunderstandings can be clarifed. Subject to social desirability, demand characteristics and experimenter effects.
Correlational analysis - Can be used when manipulation of variables would be unethical. Can't be sure the results aren't being affected by a third variable.
Content analysis - Highly relicable and easy for lots of psychologists to repeat. Material open to interpretation so may be biased results.
Case studies - Used in cases where it would be unethical to manipulate the variables. Since it tends to look at individuals there are issues with participant variables and individual differences.

Informed Consent - Participants should know the aim and procedure and should have the right to withdraw. If the participant is under 16 years old parental consent is required. It isn't required for naturalistic observations as there is no interference. An alternatives is presumptive consent (asking the general public if they would be okay with it) followed by debrief, when knowing the full study would affect the outcome.
Deception - Sometimes researchers withold information so that participants will act naturally in the study. This is allowed when there is strong scientific justification and no alternative. Participants may be given general details. Severity differs (Milgram 1963).
Protection from harm - Risk of harm shouldn't go beyond what people would face in their normal lives. Some procedures involve physical or psychological distress. Researchers may not aniticipate the harm in their studies and may need to stop the research (Zimbardo's Stanford Prison experiment).
Debriefing - This should return participants to the state they were in before the experiment. Counselling and support should be offered to any harmed participants after. Results are explained and participants are given the right to withdraw their data.
Confidentiality - Participants should be anonymous. Numbers and letters can be used instead of names. They must be warned if they are not going to be anonymous.

A significance level of 0.05 is usually used. If the probability of results being due to chance is less than or equal to 0.05 then you can accept the alternative hypothesis and reject the null.

A type one error is when you reject the null hypothesis when it was actually true. This is usually because the significance level is too big.

A type two error is when you accept the null when it was actually false. This is because the significance level was too small.

Inferential tests come up with a calculated value for the results. This value is then compared to a critical value provided in a table of values for that particular test. Depending on the test, the result is significant if the calculated value is either greater than or less than (or equal to) the critical value. Chi Squared (CS) and Spearman's Rho (SR) require a value that is greater than or equal to, while Mann Whitney U (MWU) and Wilcoxon Signed Ranks (WSR) require a value that is less than.

Relationship + Ordinal = SR --------- Difference + Ordinal + Repeated = WSR
Nominal = CS ------------------------------- Difference + Ordinal + Independent = MWU 

Qualitative data can be presented in a report or verbal summary. This often has categories to group things together, and quotations from participants.

Quantitave data can be presented in tables and graphs.

-Tables show the raw data or can also show descriptive statistics like the mean, range, and standard deviation. 

-Bar Charts are used for non-continuous data (nominal). The bars don't touch each other.

-Histograms are for continuous data with regular intervals e.g 10 second time slots. The columns touch.

-Frequency Polygons are used when showing more than one set of data on the same axes.

-Scattergraphs measure the relationship between two variables to examine a correlation. 

Abstract is a brief summary of the aim, hypothesis, method and results around 200 words long.

Introduction is a general overview of what and why you are studying it, and existing theories.

Aim and Hypotheses sentence stating the purpose of the study and what is going to be tested. Should be a null and an alternative including the IV and the DV.

Method describes how the research will be carried out. Needs to be detailed so it can be replicated. (Further explained on next card)

Results contains descriptive and inferential stats in the form of tables and graphs. Should also explain why certain graphs were used.

Discussion explains and summarises findings, and adresses implications, limitations and modifications involved. Results are related to background research and further research is suggested.

References

Appendices

(Further detail about the methods section of a report.)
Design-Research method used  
-Research design used 
-Control of variables 
-How tests were chosen 
-How ethical issues were dealt with
Participants -Number used
-Age, employment, gender etc.
-Sampling method
-How they were allocated to conditions
Procedure -Detailed step-by-step account of what happened 
-How informed consent was obtained 
-What was said and done 
-How they were debriefed
-How data was recorded
Materials -Apparatus (diagram if appropriate)
-Any tables (draw), lists, questionnaires or pictures used