Psychology Research Methods

Aim: stated as a general research area that will be investigated, looks for a difference or relationship/similarity. Usually begins 'To investigate...'

Hypothesis: prediction of what you expect to find in your study, the assumed relationship between the variables. Should be testable, operationalised and precise. 

• Experiment (lab/field/natural) = experimental hypothesis
• Not an experiment (questionnaire, interview, observation, case study) = alternative hypothesis

Non-Directional/Two-Tailed Hypothesis: states there will be a difference but not what the difference will be (no direction stated)

Directional/One-Tailed Hypothesis: states there will be a difference and what the difference will be (the direction it will go in)

Note: if previous research has been done then it must be used to form a directional hypothesis.

Null hypothesis: predicts there will be no difference or relationship/similarity. 

Researchers should always aim to support the null to avoid bias towards the experimental/alternative hypothesis, increasing objectivity. At the end of a study, either the null is supported and experimental/alternative rejected, or visa versa. 

Independent Variable (IV): variable that is changed or manipulated by the research to cause an effect on the DV. 

Dependent Variable (DV): variable that is measured to observe the effect of the change in IV. 

Both the IV and DV must be operationalised, meaning they have been put into clearly specified units. Allows more objectivity and replicability in the study, allows results to be more valid and reliable. 

Extraneous variables: extra variables that have an unwanted effect on the DV if not controlled. 

They must be controlled as much as possible to assure that it is solely the IV that has effected the DV, and a cause and effect relationship can be established. Can be controlled by standardisation. 

Three types of extraneous variable: the characteristics or behaviour of a participant, situation or investigator. 

Confounding variables: previously unidentified extraneous variables that have effected the DV. 

Reliability: how consistent a study, whether it can be repeated in exactly the same way and gather very similar results. This can be achieved by having very high levels of controls over variables. 

Validity: whether you are measuring what you think and are supposed to be measuring. This can be done by having operationalised variables. There are two types of validity:

• Internal validity: similar to the general definition stated above, the IV has been changed and the effect on the DV has been measured, with no other variables having an effect. 
• External validity: how generalisable the study is to situations outside the context of the study, is it true to real life situations? Having high levels of control would normally not be similar to a real life situation.

The aim is to have internal and external validity in a study, but this is always unlikely as more of one usually results in less of the other. 

True experiments usually manipulate the IV, have high internal validity and are useful for identifying cause and effect relationships. These include lab and field experiments. 

Laboratory experiment: IV is manipulated and participants and randomly assigned to conditions. Takes place in an artificial setting with high levels of control. 

✓ Objective and accurate due to operationalised IV, easy to replicate due to standardised procedure, precise control of variables due to artificial setting, cause and effect relationship established, high internal validity

✗ Low in external validity due to artificial setting, purposeful study may lead to investigator bias,  demand characteristics may occur if participants are aware they are being studied

Field experiment: IV is manipulated but done in every day environment of participant 

✓ High external validity due to real life situation and environment, less likelihood of demand characteristics if participants are unaware they are being studied

✗ Low in internal validity as variables cannot be controlled, not replicable, ethical issues if participants do not give consent

Both Quasi and Natural are not classsed as true experiments as they lack manipulation of the IV and random assignment to conditions. 

Natural experiment: effect of an IV which has not been manipulated by researchers as it is naturally occuring 

✓ High external validity as it is more natural to real life, less likelihood of demand characteristics

✗ Low internal validity as control of vairables is low due to natural setting, unreplicable due to low variable control

Quasi experiment: control of conditions and variables is beyond the researchers control, study a naturally occuring situation

These can be done in lab type conditions, and so share the same strengths and weaknesses. For example, high control means low external validity. However, it can also be done in more of a natural experiment setting and so it would share those strengths and weaknesses. 

✗ Allocation to the IV is not completely random, and so there may be confoudning variables which cause change to the DV

Independent Groups design:

Two separate groups of participants, take part in only one condition each. Participants shoul be allocated randomly so any differences are spread across evenly. Bigger samples sizes will reduce the effect individuals differences have on the results. 

✓ Less likelihood of demand characteristics as they have less opportunity to guess the aim of the study, less chance of order or practise effects as they are only taking part in one condiition, both conditions can be carried out at the same time possibly saving time

✗ Low internal validity due to the individuals differences which are unlikely to be fully oontrolled using random assignment, researcher must gather twice as many participants

Repeated measures/groups design:

One set of participants takes part in both conditions, one after the other.  

✓ Individual differences are not a problem as the same participants are taking part in both conditions, fewer participants are needed

✗ Demand characteristics is likely as they have more chance to guess the aim of the study, more chance of order or practise effects as they are taking part in two conditions not one

Matched Pairs design:

Similar to independent groups in that there are two groups of participants who take part in one condition each. However, these groups are paired up and matched based on relevant vairables which depend on the study. 

1. Identify relevant variables for the study and gather a large group of suitable participants.

2. Assess them on the variables, pair them up with another particpants who has similar scores on all the vairables.

3. Randomly allocate one person from the pair to condiiton A and the other person from the pair to condition B. 

✓ Less likelihood of demand characteristics as they are only taking part in one condition, less change of order or practise effects, individual differences do not effect the results

✗ Most difficult and time consuming, may need a large number of particpants to ensure they all have someone similar to them, even with the process of matched pairs individual differences may still affect results lowering internal validity

Random Allocation: used in independent groups, attempt to spread the individual differences across the conditions.

If you have a smaller number of participants:

1. Write all their names down onto individual pieces of paper, and put into a contained and shake. 

2. Draw out a participants and assign them to condition A, then another to condition B, carry on in this way until all participants have been allocated a group. 

If you have a large number of participants:

1. Allocate numbers to all the participants, use a random number generator to input all these numbers.

2. Use the generator to get half the number of participants to assign to condition A, and the other half will be assigned to condition B. 

Randomisation: either using randomisation to randomly allocate participants to different counterbalanced conditions, or using randomisation to randomly present stimuli to participants to esnure order does not serve as a confouding variable. 

Standardisation: ensuring that all participants have exactly the same experience as any other participant, and that the study can be repeated in exactly the same way by someone else. 

Standardisation should be present in the instructions given to participants, the procedure used, the stimulus matertials used and what researchers say and do during the study. 

Counterbalancing: used in repeated measures to reduce order/practise effects.

ABBA - half of the participants will complete condition A then B, and the other half will complete B then A, to balance out the order/practise effects. 

1. Put all names of participants on a list.

2. Take the first half of names and allocate them to A then B.

3. Take the second half of names and allocate them to B then A. 

Matched Stimuli: used in repeated measures to reduce practise effects.

The researcher will use two different stimuli in each of the different conditions, but match them on difficulty. This is so participants do not get better in the second condition as they have done the activity before, but also prevents the difficulty of the task from becoming an extraneous variable as it is kept the same across both conditions. 

Demand characeristics: this occurs when a participants either tries to give or is given an accidental hint as to what the purpose of the study is. This means they may change their behaviour to affect the results. This can either be in a positive way (please you) or a negative way (screw you). 

This is a problem as it means a participants behaviour becomes unnatural as they are aware they are being studied. This weakens the validity of the results. 

Investigator Effects: this is anything to do with the researcher/investigator in the study that may affect the way in which a participants behaves. This can be any type of characteristics, behaviour or feature of them. 

Some examples include: gender, age, ethnicity, appearance, tone of voice, body language, eye contact etc.

The investigator may also have an expectancy of the results they want to gain from the study, which may lead them to unconsciously encourage certain things to get those results. 

Single blind procedure: this can be used to reduce demand characteristics. This is because participants are unaware which condition of the study they are in. They are therefore unsure of which way the results are expected to go, as steps have been taken to make sure they do not know which condition they are part of. 

Double blind procedure: this can be used to reduce investigator effects as well as demand characteristics. This is because neither the investigator or the participants knows which condition the participant is in. This prevents the investigator from idadvertently giving the participants clues as to which condition they are in. 

Triple blind is also when the participant, investigator, and the person analysing the data after the study does not know which conditions the participants were in. 

Pilot Studies: this is a small-scale trial run of study which is carried out before the full study is done. This allows the researcher to check that there are no problems, and to implement a solution if there is any. They will usually check the design, the instructions, the measuring equipment etc.

A pilot study will usually save time, money and effort in the long run as the full study will only have to be done once as it will be done correctly.

Pilot studies are only done to check the procedure and method of a study, NOT to look at its ethics or results. 

Sampling: this is the selection of participants to take part in a study. These participants will become the sample in the study. Ideally, a study would include all members of its target population. However, this is extremely optimistic as it would be too time consuming and complicated to carry out a study in this way. Therefore, a researcher will use a sample, a smaller group from the target population, that will hopefully be representative of the whole target population, allowing the researcher to generalise the results to a much larger group. 

A sample can be done through one of 5 sampling methods:

1. Opportunity Sample

2. Volunteer Sample

3. Random Sample

4. Stratified Sample 

5. Systematic Sample

Opportunity sample is produced by selecting people are easily available when the study is being conducted. They just happen to be there when the study is being done. 

+ Less time consuming, no equipment needed

- Researcher has a large influence over who is chosen, may show researcher bias

Volunteer sample relies on volunteers who want to participate in the study. It is also known as a self-selected sample. It usually consists of an advertisement that participants have to respond to.

+ Less likelihood of dropouts or withdrawals as participants want to be in the study

- Likely to have a biased sample as they same sort of people normally volunteer, reduces generalisability 

Random sample occurs when every member of the target population has an equal chance of being selected. Therefore the method use must be truly random. Random allocation can be used. 

+ Less researcher bias, truly random allocation so participant variables should be spread fairly evenly across the conditions

- Sample could still be unrepresentative despite the time and effort

 Random allocation:

Allocate all people in target population a number (or just use their name). Write these down onto individual pieces of paper, then put in a hat. Pull out the number of participants that you require in your sample, and allocate to conditions as you do so. 

Stratified sample consists of participants in subgroups of frequencies that represent the frequency of that group in the population. This is done by identifying subgroups in the population, identifying how much of the population they take up (%). Then the same percentage of the sample will be participants from this subgroup of people. For example, 30% male and 70% female population. This means a sample of 50 will have 15 females and 35 males.

+ Very representative of target population, generalisable

- May need to identify all features of a population for it to be truly generalisable, also very time consuming

Systematic sample is obtained by selecting every nth person in a list (where n is any number).

+ Limits researcher bias as they have no control over who is picked, can be fairly quick and simple

- May not be representative and therefore not generalisable

Deception - researchers should not lie to participants, they should tell them the aims and what will happen in a investigation

Confidentiality - participants details should always be kepy confidential, annonimity of participants should be ensured when results are published

Consent - participants should give consent to take part in an investigation

Observation - you are only allowed to observe participants in a public place, if you wish to observe in a private place you must obtain permission

Withdrawal - the participant should be allowed to leave, and should be informed they are allowed to leave at any point, they can also withdraw their results even after the study has been done

Protection - participants should leave the investigation in the same physical, mental and emtional state that they entered the investigation in

Advice - researchers are not allowed to give advice in an area they are unqualified in

Debrief - a debriefing should take place at the end of all studies to give the participant information about why the study was done, what their part in it was, if they have been decieved and any further support they may require

Things to include in a consent form: purpose of the study, length of time required from the participants, details about the study, right to withdraw, reassurance about protection from harm, requirement to undertake any necessary tests, reassurance of confidentiality.

Things to include in a debrief: aim, independent and dependent variable, their part in the study, procedure, deception, other conditions, findings, contact details, further support information, withdrawal option. Any questions from the participants should also be answered as fully and honestly as possible. 

Presumptive consent: this is gained from people of similar backgrounds to the participants in the study. If they say they would have been willing to participate, then it may be assumed that the actual participants who are similar would be as well. However, it may be argued that it is wrong to just assume that people are the same and that they would be willing to participate. 

Prior general consent: this involves participants agreeing to be decieved, but not knowing how they will be decieved. However, as they know they are part of a study, demand characteristics may occur.

Retrospective consent: this is when participants are asked for consent after they have taken part in the study. This can be an issue if they do not consent and yet they have still taken part in the study. 

Deception: participants may have to be decieved in a study to make the study worthwhile doing and to prevent the interference of extraneous variables. In this case, they should be decieved as little as possible, and should be fully informed about why they were decieved and how in the debrief after the study.

Withdrawal: participants should be made fully aware that they can withdraw at any point before, during or after the study. They should know that their results will also be discarded if they choose to withdraw. 

Protection from harm: participants must be protected from any physical or psychological harm throughout the study, and should know that they will be protected at all costs. 

Confidentiality: all details and data about participants in the study should be kept annonymous to protect the identity of participants. This especially applies in the release of findings. 

Cost-benefit analysis: the researcher will weigh up the potential costs of the study against the potential benefits of the study. If benefits greatly outweight costs, and the costs are not significant in causing harm or any other effect to the participant, then the researcher should carry out the study. 

Case studies: longitudinal, indepth, detailed investigations carried out on a very small sample size, usually one person. It includes information about their background, medical history, life experiences, behaviour etc. The researcher may use several methods to gather data like an observation or an interview. 

+ Lots of rich, qualitative data can be gathered, high validity, can rule out indiviudual differences in personality as sample is studied over a long period of time, can be useful where other methods are inappropriate, can be used to directly challange a theory

- Case studies are completely unique and so cannot be generalised, researcher bias may occur, subjective data as it is all gathered by a researcher, a lot of case studies rely on recollection of past experiences which may be inaccurate, case studies cannot be replicated and so are not reliable

Participant: where the researcher is involved in the study, and is present in the situation that is being observed.

+ Researcher is more likely to get a true insight into the behaviour of participants, gathering rich and detailed data, high validity

- Researcher is likely to be subjective due to their expectations

Non-Participant: behaviour is recorded from a distance so researchers are not involved with the situation.

+ Likely to be more objective as they are not involved

- Data lacks detail and understanding, less validity

Overt: participants are aware that they are being watched and recorded on their behaviour. They will usually see the researcher or the cameras, and are usually aware of the purpose of the study.

+ Good ethically as they know they are being observed, have given informed consent

- Demand characteristics may occur as they are aware of the details of the study

Covert: researchers will observe without the participants knowing. This may be hidden from an observation point, hidden cameras or sometimes an undercover collegue in a group of participants.

+ Less likelihood of demand characteristics, internal validity

- Ethical issues as participants are unaware they are being studied, would have to use presumptive or prior general consent

Naturalistic: observation of behaviour in a natural setting where it is not manipulated by the researcher.

+ Behaviour is natural, external validity, true to real life

- No control of variables, low internal validity, cannot establish cause and effect

Controlled: researcher manipulates and controls variables, usually in a labratory type setting.

+ High levels of control, high internal validity, cause and effect relationship can be established

- May result in unnatural behaviour, do not reflect real life, low external validity, possible demand characteristics

Unstructured: this is a type of observation where the researcher has no set structure or checklist that they must follow. This means they can ask any questions they like and pay as much or little attention to things as they like. 

+ Detailed data, freedom to follow anything that may be of interest, rich qualitative data

- Difficult to replicate, time consuming, hard to analyse

Event sampling: this involved counting the number of times a behaviour will occur in the observation of the study. It uses behavioural categories;

A table with a list of all the possible behaviours that may occur in the observation. These should all be operationalised to make it clear exactly what the resarcher is looking for. Researchers woudl tally the amount of times they observe each behaviour and record it in the corresponding section of the table. 

+ It is unlikely that behaviour will be missed, collects quantitative data, easily analysed

- Risk of not having inter-observer reliability due to subjective judgements, may show behaviouyrs that are not specific to just one category on the table

Time Sampling: researcher will decide an interval of time inbetween each recording of behaviour e.g. 30 seconds, and behaviour will be recorded at the exact moment between these time periods. 

They may use this as well as a behavioural categories table, so the data is more quantitative than qualitative. 

+ Observers have time to record what they see, easily replicable

- May miss behaviour inbetween recordings, not representative

Questionnaire: list of pre-meditated questions for participants to respond to. They may be multiple choice, or may have space to answer more freely. They can either be given to participants with the researcher present e.g. face to face, over the phone, or they may be done without the researcher present e.g. on a website, email. 

+ Time efficient compared to interviews, get a larger number of participants to take part, more cost-effective, may result in more honest answers, likely to be more natural answers, replicated easily, gather both types of data, low investigator effects

- Response rates can be poor, can only gather very limited (usually quantitative) data, wording may skew interpretation of questions, only certain types of people will return questionnaires (biased), hard to generalise, may have leading questions, 

How to write a good questionnnaire:

Make it relevant to the aim of the study, easy questionnaies at the start, put in questions that disguise the aim of the study, ensure clarity of what is being asked, no ambiguity, avoid bias within questions, don't suggest one answer is preferred over another to avoid social desirability, make it fairly short, choose relevant measurement scales, do a possible pilot study

Open questions: these type of questions do not have set answers and allow participants to respond freely with no restriction.

+ More flexibility in the way they respond, can explain viewpoints, useful when full range of answers is not known, can allow lots of rich and detailed data

- Generates qualitative data which may be hard to analyse, less likelihood of being completed as they take more time and effort, unspecific questions could lead to answers that irrelevant to what they researcher wants to know

Closed questions: fixed questions that have options for the answers and participants must choose one that most suits them. 

+ Generate quantitative data which can analysed for patterns, useful for answers that need to be discrete or do not require explanations from participants, more likely to be completed

- Must have a full range of possible answers, invalid data if participants answer is not available, lack of engagament needed as they could just tick any answer to get it finished

Likert scale: a type of scale used in psychology questionnaires where participants respond to a question (usually closed) with a set number of answers. They are often used to assess attitudes, personality and behaviours. Usually uses a series of agree or disagree options. For example:

Strongly Agree, Agree, Neither, Disagree, Strongly Disagree

+ Easily understood, collects quantitative data, easy to analyse, quick and efficient

- Only gives a limitied number of options, cannot measure true opinion of all participants, participants often sit on the fence and avoid the extreme answers 

Interviews: researchers and participants engage in a face-to-face conversation where participants will respond verbally to quesitons asked by the researcher. It is similar to questionnaire, however the participant is being asked face-to-face and their answers will be recorded either by written notes, a visual recording or an auditory recording.

+ Rapport can be built up meaning participant may be more likely to trust the researcher meaning they should be more open and honest, more validity, socially sensitive topics can be addressed, good source of qualitative data, lots of detail, can ask follow up questions if relevant

- Subject to researcher bias, investigator effects/demand characteristics, only effective if the participant is open and honest, and if interviewer is skilled and trained

Designing an interview: may have 'filler' questions to disguise the purpose of the study, be prepared - record the interview to refer back to later, be aware of non-verbal communication from the interviewer or interviewee, ensure the interviewer is skilled and trained

Structured interviews: researchers use a set of predetermined questions, that they must not deviate from.

        + Data can be analysed and compared more easily as they all answer the same questions

        - Possibly more time consuming and costly than a questionnaire but offer little more, may not get as much rich and detailed data due to the set and ridgid structure of questions

Unstructured interviews: no set questions, just a topic focus that should be addressed, flexibility and freedom to lead the conversation further and allow the participant to expand.

        + Lots of in-depth detailed information

        - Difficult to analyse as different participants may have answered different questions, requires a skilled and trained interviewer (more so than structured) as they must be able to engage with participants

Semi-structured interviews: there are a set list of questions but the researcher can deviate from these can take the interview in whichever direction they feel is suitabl (half structured, half unstructured).

        + Can lead onto suitable conversations based on the responses of participants, still have set questions to allow structure to the interview 

Correlation: looks at the relationship between co-variables, without any manipulation of variables. 

They are different to experiment in that they do not look for a difference or a cause-effect relationship, like an experiment does. It only looks for a relationship. They will illustrate the strength and direction of an association between two co-variables. 

+ Useful for a starting point for research, quick and economical, more suitable when manipulation of the IV is not an option, can be effective predictions, test for significance to possibly generalise

- No cause-effect is established, may be other variables having an effect, may be misused/misinterpreted by the media, only useful for linear & simple relationships 

If a relationship between co-variables is found, the researcher must decide how to interpret the relationship. They may say that there is a direct cause-effect relationship between the variables, or that there is a third variable that is causing the relationship. They may also conclude that it is a spurious correlation, meaning it has happened purely by chance and the co-variables are not related in any way.

Correlations should be presented on a scatter graph so the direction/strength can be identified. 

A correlation should have a strength and a direction to give more detail about the relationship between the two co-variables. 

Direction: this will state which way each co-variable goes in relation to the other

                Positive - as one variable increases, the other variable increases

                Negative - as one variable increases, the other variable decreases

Strength: this will state the strength of the link between the co-variables                                                                                      (using a correlation co-efficient between 1 and -1)

1.0 - 0.8 = very strong correlation                                         correlation > 0 = positive

0.8 - 0.5 = strong correlation                                                 correlation < 0 = negative 

0.5 - 0.3 = weak correlation

0.3 - 0 - very weak correlation

Qualitative: non-numerical, language based data to do with opinions, experiences and judgements. Deals with descriptions of data that is observed rather than measured. Usually rich and detailed data that is to do with individual experiences. Tends to be subjective and can be time consuming and difficult to analyse. 

Quantitative: numerical data that can be statistically analysed, usually data that can be measured.It is usually quick and easy to collect and analyse, and is objective. Sometimes tries to generalise but often lacks enough detail to be useful.

Primary data: information observed or collected directly from first-hand experience (collected by the researcher for the purpose of the study being carried out). 

+ Information collected is specific to the study and designed to fit the aim and hypothesis

- Time consuming and often expensive

Secondary data: collected for a purpose other than for the current study. sources include government statistics, previous study research, newspaper reports etc. 

+ Easier and cheaper to use the research someone has already done, may be significant if statistical testing has been done

- May not fit the specific needs of the current study, may have already been interpreted/manipulated and so may be biased

Meta-Analysis: primary data from other previous studies is re-analysed to create new 'quantitative metadata'. 

+ Useful for topics that have been extensively researched, can identify common trends across several studies

- Criteria has to be very precise and strict which can limit the number of possible studies that can be used, relies on primary data being of very good quality

Nominal: categories of data where items will fall into certain a category based on certain characteristics

e.g. gender of students in a class (male or female)

This data will show frequencies of the different categories which can be put into a table, bar chart, pie chart etc. 

Ordinal (ranked): ranked data to see the data values in relation to each other

e.g. who came 1st, 2nd, 3rd etc in a race

Can be put into nominal (categories) based on the ranks. Can calculate a median or range. 

Interval (ratio): stated in a rank order including the precise intervals between values

e.g. in a race Andy 1st (10seconds) Bob 2nd (11 seconds) Charles 3rd (13 seconds) 

Put into nominal or ordinal, or produce measures of central tendency or dispersion. 

Mean: statistical average of all data. Found by adding all values up and diving by the number of items. 

+ Most sensitive as it accounts for all items

- Can be distorted by extreme values, often not one of the actual items, may involved decimals

Median: middle value when all items are in ascending order. 

+ Not affected by extreme values, easy to calculate, more representative than the mean with small sample sets

- Distored by small sample or repeat values, not as sensitive as the mean

Mode: most popular value

+ Not influenced by extreme values, useful to identify most popular 

- Sometimes more than one mode in a set of values, shows nothing about other scores

Range: spread of values (difference between smallest and biggest value). Found by taking smallest value away from largest.

+ Easy to calculate, shows range of data values

- Easily distorted by extreme values, gives no information about other values in a data set

Standard Deviation: measure of spread of values around the mean (average distance between all values and mean value). 

+ Most powerful as it takes all values into account, sensitive measure

- More difficult to calculate than the range, easily distorted by extreme scores as it is based around the mean

If SD is high, the data points are spread far away from the mean. If SD is low, the data points are clustered around the mean and are seen as more consistent. 

Percentages

Number in category divided by total number in category multiplied by 100. 

Converting percentage to decimal

Percentage: AB%                                        Decimal: 0.AB

Normal Distribution: mean is fairly central to all the data points, and data is generally distributed evenly. It may be seen as symmetrical when displayed on a graph, and is usually described as 'mound-shaped'. The mean and median are often very close or the same. 

Positive Skew: majority of data is clustered up towards the left hand side of the graph. This is because the mean is larger than the median. 

Negative Skew: majority of data is clustered up towards right hand side of the graph. This is because the mean is smaller than the median.  

Statistical tests find out how likely our results will reflect not just our sample but the whole population. They do this by stating whether the results are significant or not.

The accepted level of probability in psychology is 5% (P=0.05 where P is chance that results were caused by chance and not by the IV affecting the DV).

If P<0.05 then results are classed as SIGNIFICANT - the experimental hypothesis is accepted and the null hypothesis is rejected.

If P>0.05 then results are classed as NOT SIGNIFICANT - the null hypothesis is accepted and the experimental hypothesis is rejected.

Significance indicates that the results can be generalised to the rest of the population.

The sign test is used in these conditions: looking for a difference, repeated measures/matched pairs, ordinal/nominal data.

How to conduct the sign test:

1. Convert the data into nominal data by working out the difference between the data values. State whether the difference is positive or negaive by assigning +'s or -'s.

2. Count the number of +'s and number of -'s. The least frequent will be used as the S value.

3. Add together all the value where there was a difference (add total +'s and total -'s) and define as N value.

4. Identify the critical value using the table. Compare with calculated number S.

5. If S  ≤  critical value , then results are signifcant. So is S is less than or equal to the critical value then results are classed as significant and the experimental hypothesis can be accepted.

A significance statement must be written. It must include whether the results are significant and why, the calcualted value (S), the value of N, the critical value (from the table), and finally which hypothesis is accepted/rejected and the conclusion because of this.

Peer review: a process that takes place before studies/research is published to ensure it is of a high quality, contributes effectively, and is accurately presented.

This process is seen as very important it ensures the research is all correct and will have a positive impact rather than negative.

+ Promotes and maintains high standards in research, prevents scientific fraud, promotes scientific processes through the development of knowledge and research

- Annonimity may not be maintained which could have negative consequences, there is a risk for bias from individuals in order to further their own career, findings that challange existing views may not be published due to them being 'not significant'