Research design (second half)

Descriptive stats

summarizes what is observed

Binary measurement

2 distinct things

Nominal measurement

2+ distinct things

Ordinal measurement

order/rankings to categories

Interval measurement

scores that have equal intervals on the scale with no true 0

Ratio measurement

can make meaningful ratios with a true 0

Mean usage

data has no extreme scores and is uncategorical

Median usage

data has extreme scores and no distortion to the average

Mode usage

data is categorical and values can only fit into 1 class

Range

measures difference between highest and lowest values in the dataset

Standard deviation

how far each data point is from the mean on average

• Large SD= inconsistent

• Smaller SD= consistent

Variance

average squared deviance from the mean

• High = separated data

• Low = closer data

Skewness

measure of symmetry and asymmetry of the data

• none = normal curve and even distribution

• Positive = tail to the right, data on left side

• Negative = tail to the left, data on right side

• -0.5 to 0.5 = symmetrical

• -1 to 1 = moderate

• < -1 to > 1 = high

Kurtosis

how peaked or flat a distribution appears

• normal = 0

• platy = < 0; separated, flatter, variability

• lepto = > 0; close together, peaked, less variability

Inferential stats

data generalizes from sample to population

Z-scores

changes scores into a unitless, consistent score (score - mean)/ SD

• ± 1,2,3 → distance from the average

Null hypothesis

The idea that the treatment being tested has no impact; H1 means that new methods improves accuracy

Type 1 hypothesis error

Rejecting the null hypothesis when it is actually true (ex. ref calls foul that didn’t happen)

Type 2 hypothesis error

Failing to reject the null hypothesis when it is actually false (ex. ref misses an actual foul)

Stat Power

chances of finding a true effect; larger sample size is vital in finding change

• Higher = more confidence in results and reliable conclusions

• Lower = higher chance of missing true effect

T-test stats

compares two groups

F-test stats

compares differences in 2+ groups; ANOVA

X² test stats

compares frequencies; Chi squared

Significance level

the risk in not being fully confident that what is observed in an experiment is due to the treatment/what’s being tested

P-value

a level of probability that the effects seen are due to the treatment

• <.05 = stat. significant

• >.05 = pract. significant

Confidence interval

gives a range likely to contain the population boundaries; shows precision using a percentage and range

Effect size

quantifies how meaningful a result is; uses cohen’s d

Cohen’s guidelines

• Small: 0.2 to 0.49

• Moderate: 0.5 to 0.79

• Large: greater than 0.8