experimental designs

four types of dependent variables?

• nominal

• interval

• ordinal

• ratio

what is mean, median, mode?

• mean is average of all numbers in data set

• median is 50th percentile, middle number in data

• mode is number appearing most often in data set

SD? standard error?

• measure of variability

• based on SD, how different the population mean is likely to be from sample mean

correlation

• amount of relation btw X and Y

when is something statistically signif effect?

p<0.05

p value

• probabiilty that nay particular outcome would have arisen by chance

• high p value means data is likely with a true null (0.78)

• low p value means data is unlikely a true null (0.042)

• btw 0 and 1 for chance

independent vairables

• what is manipulated

• treatment

dependent variable

• what is measured/expected to change

• performance

what is true experiment

• group randomization

• can have control group with no treamtnet

quasi experiment

• sometimes multiple groups

• can be selective not random

• not necessarily involving a control/comparison group

scientific method?

• observe

• question

• hypothesis

• procedure

• experiment/data

• results

• conclusion

what are we doing in an experiment?

• manipulate variable to see effect on other/more variables

level?

• within a factor

• eg 3 diff duration conditions=3 levels of the factor duration of intervention

factor?

• the controlled IV

• eg duration of intervention

design types

R, O, T, G

single factor

• one treatment (IV, ex with puppets or without)

• several levels

• t test

• anova

multi factor design

• more than one IV

• multiple groups

• multiple conditions

factorial design

• more than one IV

• multiple groups

• multiple conditions

data analysis

• qualitative, themes

• quantitative, inferential stats/descriptive stats/correlations/regressions

inferential stats

• inductive inference with prob

• quantification of uncertainty

two types of error

• type I (false pos, maybe alpha level set too high .10)

• type II (false neg, maybe alpha level set too low, 0.01)

effect size

• dif btw mean outcome in two groups

• cohens D, odds ratio or correlation

Confidence interval

Plausible values for the true population average or treatment
effect, given the observed data.
• Level often 95%: If you repeat this experiment 100 times, the
means of each sample will tend to fall within the given range 95
times out of 100.