Hypotheses, Sampling and Design

First research methods post today! More will follow as I consolidate the rest of the unit.


Black: Description
Red: Negative evaluation of a design or technique.
Blue: Positive evaluation of a design or technique.
Purple: My notes/hints/tips

Experimental Designs

An independent groups design has different groups of participants randomly allocated to each condition of the independent variable. For example, an experiment with two conditions of the IV and a control group would have three different groups of participants - one for each condition. Random allocation is used to prevent researcher bias from favouring one IV condition over another and consciously or unconsciously selecting the group to reflect this favour.

No order effects – the same participant does not participate multiple times, so there are no carried over effects of practice, boredom or fatigue.

Lower chance of demand characteristics – if people are only in one condition of the IV, they are less likely to attempt to guess the aim of the study and change their behaviour to either help or hinder the researcher.

The same apparatus can be used in both conditions – since participants only carry out one IV condition, the experimental stimulus will be unfamiliar to them – they will not have encountered it before.

However, more participants required than for the repeated measures design, so it’s more time consuming.

Also, individual differences between participants in each group may affect results – variations in age, sex, social background and ability that reduce validity.

In a repeated measures design, the same participants take part in each condition of the independent variable. This means that each condition of the experiment includes the same group of participants. 

Fewer participants are needed as the same group part in all conditions – saving time in gathering the sample.

There are no individual differences between groups that could affect the results – the individual differences are only internal, between participants in the same group.

Order effects may reduce the validity of the results, the order of the conditions affecting participants’ behaviour.  Performance in the second trial may be different due to practice, fatigue, and boredom. Counterbalancing can be used to spread order effects equally between all the IV conditions – participants are randomly assigned to complete IV conditions in different orders  - one group does (A) then (B), the other does (B) then (A), helping to offset order effects.

Demand characteristics are also more likely in this design – if people participate in all the conditions of the IV, they are more likely to guess the aim of the study and adjust their behaviour to help or hinder the researcher - both of which damage validity.

The researcher can't use the same apparatus twice – participants will have already encountered the stimulus in the first condition, so it cannot be reused in the second.

In a matched pairs design, participants are paired up with someone with very similar relevant abilities and characteristics, and are then randomly assigned to IV conditions or the control group. Ideally, somebody in one IV condition would be paired to their identical twin in the other condition - but this is rarely practical.

This design avoids order effects, so counterbalancing is not necessary.

Reduces individual differences between the conditions, because the participants are paired up so each has somebody in the other IV condition with similar attributes and characteristics.

The same apparatus can be used in both conditions – since participants only carry out one IV condition, the stimulus will be unfamiliar to them – they will not have encountered it before.

Demand characteristics are less likely, as each participant only participates in one condition of the IV so there is less chance of them trying to guess the aim and changing behaviour.

Sampling Techniques

An opportunity sample consists of simply selecting people as participants who are available and willing at the required time – for example, just going up to people on the street and asking them to participate in your research.

An opportunity sample is quick, convenient, and practical – allowing for the selection of many participants fairly easily just by using those who are available and willing.

Very unrepresentative sample as you only get a specific social demographic – the kind of people likely to be around at the time and place of sample gathering.

Often biased by the researcher, who will likely choose people who present as “helpful” – rather than those representative of a range of personalities and temperaments.

In a random sample, every member of the target population has an equal chance of being selected as a member of the sample population – using methods such as pulling names from a hat.

For very large samples, this provides the best chance of an unbiased and representative sample – theoretically the random nature of selection will not lead to any biases, as everyone has an equal chance of selection.

No researcher bias, as the researchers have no control over whom from the target population is selected as a participant.

 

For large populations, it is time-consuming and difficult to create a list of every individual in the target population, so a less representative but easier sample such as a volunteer or opportunity sample might be more practical.

 

A volunteer sample works through self-selection – individuals who have chosen to be involved in a study form the sample population, for example, people who responded to a newspaper advertisement asking for participants in a study will be chosen to participate.

Relatively convenient and ethical if it leads to informed consent – a large sample of participants can be gathered very quickly for a large-scale study such as a questionnaire or survey.

The researcher can advertise for specific traits and demographics in participants that they desire, if relevant, – e.g. Milgram’s use of a volunteer sample to select those to form a specific sample population of males between age 18 and 60, simulating the gender and age demographics of Nazi soldiers in WWII.

Unrepresentative, as it is biased towards the kind of person who will respond to advertisements and volunteer their time – biased towards a specific helpful kind of personality and temperament that is not representative of the general population.

Hypotheses

Hypotheses are testable, provable and falsifiable statements that predict the outcome of a study.

A directional hypothesis suggests that different IV conditions will affect the DV, and the direction of the effect – which way the DV will influence the IV. For example, "participants who have recently eaten carrots will be able to correctly identify more objects in a dark room than those who have not recently eaten carrots.

A non-directional hypothesis suggests that different IV conditions will affect the DV, but does not suggest a direction. For example, "there will be a difference in the number of objects in a dark room correctly identified, between participants who have recently eaten carrots, and those who have not recently eaten carrots."

A null hypothesis suggests that the IV will not affect the DV, and any difference in results will be due to chance or experimental error. For example, "there will be no difference in the number of objects in a dark room correctly identified, between participants who have recently eaten carrots, and those who have not recently eaten carrots."

Before being used to a form a hypothesis, variables must be operationalized - being made quantifiable and measurable. For example, the operationalised IV in the above is "eaten carrots recently: yes or no", and the operationalised DV is "number of objects correctly identified in a dark room."
An alternative hypothesis is any hypothesis other than the null.