RMS 1

null hypotheses

the treatment does not have an effect (cannot be proven)

experimental hypothesis

the treatment has an effect (IV → DV)

control group

• “no treatment” or neutral condition

• placebo: control group that’s exposed to an inert condition

experimental group

experiences a level(s) of the manipulation

blind study

masked; participants know which group they’re in, but the observers don’t

double-blind study

neither the participants nor the researchers who evaluate them know who is in which group (best)

independent variable

levels are managed by the researcher (treatment/condition/manipulation/cause)

dependent variable

scores/values that are measured by the researcher (effect)

what is power?

• the ability to detect differences or a relationship between groups when it actually exists

• the more powerful a study is, the more likely that the relationship/differences will be detected

how to increase power?

• increase number of participants (let random error balance out)

• build up treatment effect

• use a previous study to see about what the effect size should be (power analysis)

• reduce the effect of random error

  • standardize procedures and use reliable measures

  • use a homogenous group of participants (limit within-group variability)

  • code data carefully

requirement for an experiment

independent variable is manipulates, dependent variable is measured

random sampling vs random assignment

random sampling (from the population) affects external validity, while random assignment (from the sample) affects internal validity

type I error

saying there is a different when there isn’t (“there is none”)

type II error

saying there is no difference when there is, fix by increasing power (“more power to you”)

significant vs nonsignificant results

• significant: confident beyond reasonable doubt that the difference is due to treatment

• nonsignificant: can’t reject the null hypothesis, but the difference could still be due to treatment

relationship between p-values, power, type I error, and type II error

• p-value (alpha-level) is what you use to determine if result are significant- defines type I error probability

• decreasing p-value increases power and type II error probability

advantages of multiple group experiments

• comparing more than 2 kinds of treatment (may help discover significant relationships, map the functional relationship)

• comparing more than 2 treatments against each other and against no treatment (increase construct validity, empty vs placebo control groups)

• better ability to estimate the effects of different amounts (levels) of a treatment

• can help limit confounds

disadvantages of multiple group experiments

• need more participants

• more expensive

empty control group

• some subjects receive no treatment at all

• get baseline levels

• can’t see if there’s a mental effect → don’t know why the treatment works

placebo control group

• some subjects receive a “treatment” that is not expected to make a difference in the dependent variable

• see the mental effect

• no real control → no baseline levels

functional relationships + types

• the pattern of relationship between different amounts of the treatment and the dependent variable

• linear: at least 2 levels, line goes up with more treatment, 0 bends

• quadratic: at least 3 levels, looks like a parabola, 1 bend

• cubic: at least 4 levels, squiggle, 2 bends

f-test

• use for a factorial design instead of a t-test (otherwise risk type I -false positive- error)

• rarely expect values less than 1

• if only chance is operating, expect value to be 1

f-statistic

• (between-group variance)/(within-group variance)

• (mean squares treatment)/(mean squares error)

• (error + treatment)/(error)

within-group variance is

error variance

between-group variance is

error variance + treatment

error variance is

within-groups variance (caused by individual differences)

advantages of 2×2 factorial

• can compare a single treatment to a multi-stage placebo/control

• can see interaction effects

how many overall and simple effects in a 2×2 factorial?

2 overall and 4 simple

how to calculate overall main effects

by averaging a treatment’s simple effects (the average effect of a varying factor)

how to describe simple main effects

“the simple effect of ___ within the ___ condition is ___.”

how to recognize and interaction

simple effects are different across levels; lines on a graph cross

how to form matched pairs

subjects are similar to one another on one (or more) dimension, randomly assign one member of each pair to the treatment condition and the other ot the control condition

how are matched experiments better than other between-subjects designs?

• between-subject designs are wasteful (require more participants) and not powerful (treatment effects are hidden because groups are different)

• matched-pairs designs fix this error by increasing power: we reduce the chances that groups will differ due to error alone

within-subjects design

• each subject receives all treatments

• each subject is measured after each treatment

advantages of within-subjects design

• power

  • it virtually eliminates pre-existing differences between groups

  • every subject is in each group, so pre-treatment scores are identical

  • not we can more easily detect meaningful differences between treatment groups

disadvantage of matched-groups designs

may be more susceptible to order effects

effect of using within-subjects designs on…

• power: increases

• external validity: increases

• number of subjects required: decreases

• effects of random error: decreases