parameter
any characteristic/calculation of a population
statistic
- any calculation/calculation of the sample
- used to estimate population parameter
- differences between sample groups are called sampling vairablity
standard error of mean
- looks at the difference of each sample mean and the pop mean
- greater means data has lots of variability
- calculated using the population and sample size
- SEM error bars used when sure of the pop mean
standard deviation
looks at the difference between observations and the sample mean
random variables
any measurable/categorized characteristic where values/outcomes depend on chance, visualized by probability distributions
binomial distribution
discrete, dichotomous, mutually exclusive with a fixed number of independent trials
types of probability distribution
binomial distribution
poisson distribution
guassian (normal) distribution
poisson distribution
binomial event but events occur infrequently with no restrictions on the number of trails
gaussian (normal) distribution
continuous variables with infinite number of outcomes
- pop mean = 0
- pop SD = 1
z score
- represents the number of SD between random variables and the pop mean
- calculated using variable, pop mean and pop SD
- used to calculate the probability of variables occurring in certain ranges
empirical rule
- the probabilities associated with each range for normal distribution
- occurs one, two or three SD of mean
central limit theorem
as sample size increases all sampling distributions will approximate the normal distribution
- mean approximated pop mean
SD = standard error
estimation
estimates a pop parameter using sample statistics and probablity
hypothesis testing
making decisions about a value of a pop parameter using a sample of data and probablity
interval estimation
range of values likely with the population parameter
- CI, 95% confidence true pop falls within range
point estimation
single value (sample mean) used to estimate pop parameter
confidence interval
95% CI falls within +/- 2 SD
hypothesis testing decisions
fail to reject the null - not enough evidence to suggest differences (p greater than 0.05)
reject the null - evidence suggests differences (p less than 0.05)
null hypothesis (H0)
status quo, no difference
alternative (H1)
differences, can be one tailed (upper or lower) or two tailed
significance level
the risk willing to take that findings are incorrect (5% chance of incorrect)
rejection region limit
two tailed 5% split (2.5% upper and lower), one tailed 5% all on one side
type I and II error
type I null is true but you reject the null, type II null is false but you accept the null
