research and stats i w3: associational research

Research designs

• the way we structure a study

• Ex: number of sample and conditions, procedure for participants, how to test the constructs (ex, survey, heart rate etc)

Associational designs

• designs to look at associations between variables. Ex: variable 1 changes, what happens to ariable 2?

• Ex: studying association to good grades, income happiness, stress & hours of sleep

• Not looking at what causes what, just correlation

Correlations

• measured on -1 to +1

• Size of value: closer to 1 or -1, the stronger the association is between two variables

• Cohen: 0.1 (small correlation) 0.3 (medium) 0.5 and up (strong)

• Hard to get strong correlation due to other possible factors

• 0 = no association

• - sign = negative association

Positive correlation

• one variable goes up, other variable goes up

Negative correlation

• one value goes up, other goes down

Interpreting correlations

• shape (in graph)

• Sign/direction (pos/neg)

• Size/magnitude (strength)

• Statistical signifiance

(Lurking) third variable

• other (unmeasured) variable impacting the two variables we’re measuring

Causal pathways

• How the correlated variables interact with each other. A —› B or A ‹—› B or C —› A + B

• correlational designs do not demonstrate causation. Causal pathways are NOT involved in correlational designs!

Spurious correlations

• associations between variables that don’t actually make sense, just random

• Associates degrees awarded in business and management correlation to scumbag steve meme

• Strong relation, but not causal

• Therefore good to have a theory before you investigate. Constrained all possible explanations

Between groups designs

• AKA independent groups designs

• Looking at two groups o people and looking at the differences between them

• Each group experiences different independent variables (IV)

• Non-repeated measures of one person

• Ex: exercise and stress. 1 group exercises a lot, other doesn’t exercise a lot

• k = the levels of IV, the different conditions. Ex: if you have 2 different conditions, k = 2

• Can have several k in one IV (ex: only use phone on weekends, only use phone on weekdays. K = 2)

• Ex: age groups 15-17 (k = 3)

Forms of between group designs

• diff research settings

• Diff group formations

Random groups

• randomly assigning poeple to diff conditions

Natural groups

• independent variable has not been planned or manipulated by researcher. Naturally occurred. ????

Matched group

• reseacher picks their group specifically. Ex: finding ppl who are dyslexic etc

Laboratory experiment

• artifical situation, devidd by researcher

• Highly controlled

• Ex: Milgram

Field experiments

• conducted in the real world

• Less control of external factors

• More external validity than laboratory experiments

Trade-offs

• diff designs have certain skills and drawbacks

• Consider: what important to the aims? Practical issues? Confounding effects? Ethics? (Deception, etc)

Within subject designs

• Aka ”dependent group designs” or ”within groups designs”

• Repeated measures - several data points from the same person

• Changes: different timepoints or different conditions, each participant provides data at all levels (k) of the IV

Participant vs subjects

• we dont like subjects bc participant is ”subject” to the researcher. Not really great

Longitudinal designs

• common in WSD. Come back several time throughout life

Interventions or training studies

• test before intervention

• Test shortly after

• Could even test again months later

using a training or intervention to see if it works/has an impact

Comparing conditions

• test the same participants under different conditions

• WSD

• Could see how fast ppl recognize smiling or neutral faces. Comparing these two IVs

Within subjets design: k

• k is conditions. Different conditions/levels

• These can be within or between groups

• in WSD, it can be different points in time which are different levels for example

Why use in subjects designs?

• require fewer participants

• Increase sensitivity of test (less random variation, comparing one to one)

• More efficient and convenient (less recruitment, scheduling, testing time)

• Studies changes over time (longitudinal studies or intervention studies)

Between groups: issues

• if we’re measuring one variable between these groups, it does not consider lurking variables.

• Ex: maybe smokers are older? Doesn’t consider culture? Smoking vs not smoking may not actually find a clear answer

• Need to be sure that sample size is big enough to b able to distinguish variation

• fewer differences over time in WSD