PS1107 - Psychological research skills 2

Independent variable

• Specific factor that the experimenter manipulates or changes to see its effect

• For example in a study on caffeine and memory the amount of caffeine given to the participants

Dependent Variable

• the outcome variable that researchers measure

• It is assumed to be affected by the changes made to the IV

• Such as performance score on a caffeine study

Levels of an independent variable

• the different values or conditions used for the independent variables

• E.g 0mg, 50mg and 100mg for caffeine study - IV has three levels

Extraneous variables

• extra factors that have the potential to affect the dependent variable

Confounding variable

• when an extraneous variable changes systemically along with an independent variable

• Provides a alternative explanation for results

Does correlation equal causation?

• no - just because two things change together does not mean one caused the other

Why is establishing cause and effect important?

• helps us know exactly why things happen which allows for better real world interventions

• E.g if we know sleep causes better grades we can encourage more sleep to improve student performance

What three things have in a ‘true experiment’

• manipulates an independent variable

• Holds all other variables constant - to control for extraneous factors

• Measures the change in the dependent variable

Control group

• provides a baseline measure of what happens without the specific treatment or intervention

• Allows researchers to compare the treated groups results to a normal state

Placebo group

• receives an inert treatment with no active ingredients

• Helps researchers see if the results are caused by the participants expectations of an effect rather than the treatment itself

Independent samples design

• involves randomly allocating participants to different groups where each person only takes part in one condition

Weakness of independent samples

• participant variables (individual differences)

• Different people are in each group differences in results might be due to the people themselves rather than he IV

Repeated measures design

• uses the same participants for every condition of the experiment

• This eliminates differences between participants as a factor

What are order effects and how do you fix them

• occurs when the sequence of tasks affects the results (e.g participants get better with practice or worse due to boredom)

• Can be fixed with counterbalancing which means half the participants do condition a then b while the other half does b then a

Matched pairs design

• method where participants are paired up based on specific characteristics (like IQ or age) that might interfere with results ensuring the groups are equal before the experiment starts

Quasi-Experiment

• Used when it is impossible or unethical to randomly assign participants to groups

• Includes studying differences based on age, gender or pre-existing medical conditions like limb amputation

Pro’s of online experiments

• easy to recruit large diverse groups quickly

• Can be done when face to face testing is impossible

Con’s of online experiments

• no control over the participants environment

• Not practical for physical tasks like exercise or tasting food

• Potential technical glitches

who developed the t-test

• William Sealy Gosset - 1980

• While he worked at Guinness brewery

One-sample t-test

• whether a single group differs from a known value

Independent samples t-test

• Whether two separate groups differs from each other

Paired samples t-test

• Whether there is a significant difference between paired measurements

Why is degrees of freedom important in a t-test

• determines how well the t-statistic approximates a normal distribution

• Generally as the sample size increases the critical value of t gets smaller for a paired samples t-test the formula is DF= n- 1

What assumptions must be met to use a parametric test like the t-test

• normality - data in each group should be normally distributed

• Equal variance - groups should ave approximately equal variance

• Independence - data should be randomly and independently sampled

• No outliers - there should be no extreme outliers

• Measurement level - data should be at least at the interval level

Advantages of a repeated measures design

• reduces individual differences and requires fewer participants

Disadvantages of repeated measures design

• it can lead to order effects - can be managed through counterbalancing

H0 for paired samples t-test

• null hypothesis - assumes the true mean difference between the paired samples is zero

H1 for paired samples t-test

• alternative hypothesis - assumes there is a significant effect of the independent variable on the dependent variable

Standard benchmarks for Cohen’s d

• 0.2 - small effect

• 0.5 - moderate effect

• 0.8. - large effect

How Should values and effect sizes be formatted in a report

• because p-value cannot exceed. 1.00 you leave off the initial zero e.g write .05 instead of 0.05

• Because d can be larger than 1.00 you must include the zero before the decimal point

When do you use one-sample T-test

• when you want to compare the mean of one single group to a known or hypothesised population mean. See if there is a significant difference

• There is no comparison being made between different groups in this test

Main assumptions for a one-sample t-test

• the variable is measured on a continuous scale

• Data points are independent - no relationship between them

• The data is approximately normally distributed

• There are no significant outliers

What are the null and alternative hypotheses for a one sample t-test

• H0 - the population mean equals the specified mean value

• H1 - the population mean is different from the specified mean value

How do you calculate degrees of freedom for one sample t-test

• the formula is DF= n-1 where n is the total number of participants in the sample

Independent samples t-test

• this test determines if there is a statistically significant difference between the means of two independent groups

Pros of independent measures design

• avoids order effects because participants only take part in one condition

Cons of independent measures design

• it requires a larger sample size and individual differences between the particpants in each group may effect the results

Main assumptions for an independent samples t-test

• Independent groups - each participant provides only one data point for one group

• Continuous and normal - the dependent variable is continuous and approximately normally distributed

• No outliers - there are no significant outliers in the data

• Equal variances - the variance in each group should be equal

What is Levene’s test and why does it matter

• checks the assumption that the variances of two groups are equal

• If this test is significant you must report the Welch’s adjusted t-statistic instead of the standard students t-test

How to calculate degrees of freedom for an independent sample

• DF = n-2

non-parametric test

• is a “distribution-free” test meaning it does not require your data to follow a perfect bell curve

when should the Mann Whitney U test be used

• when you want to compare two seperate and unrelated groups - also called independent samples

what is a simple way to remember what the Mann Whitney test is for

• U stands for unrelated groups

ordinal data

• data that is put in to an order or rank rather than using exact measurements

how do you handle tied scores when ranking data

• you give the tied scores the middle point (average) of the positions they would have taken

to find a significant difference should your U value be large or small

• it must be equal to or smaller than the “critical value” found in a statistcal table

After calculating a "U" value for both groups (UA​ and UB​), which one is your final answer?

• the smaller of the two values is used as the final U statistic

what does a small U value actually tell you about your groups

• it means there is very little overlap between the two groups suggesting that they are significantly different

which average is the best to report for the Mann Whitney U test

• the median is recommended because it is more meaningful for ranked data than the mean

what is the effect size called for the Mann Whitney U test

• called the rank-biserial correlation