Normality is a common assumption of inferential Power (hit): probability of identifying an outcome
statistics. Models can be used to calculated likelihood of when there is one (when null is false, if null is true,
outcomes. It has a mean, standard deviation, is there is no power), increase by:
unimodal, symmetric and mesokurtic. Increase effect size
On the table, the value of one specific outcome is 0 Increase n (CLT)
because conceptually, there are infinite possibilities, so Use 2-tailed to have power on both sides
one value would be so small.
Increase alpha (but not preferable as that
increases Type I error probability)
Linear transformations of mean state that whatever you
do to original, you do to the other. Linear Central Limit Theorem: for a random sample of
transformations of variance state that if you add or observations from any distribution with a finite mean
subtract, you don’t change variance, but if you multiply
and a finite variance, the mean of the observations will
or divide, you multiply or divide by the square. follow a normal distribution. This is the main reason of
widespread use of statistics based on the normal
The “68-95-99” Rule says that in normal distributions, distribution.
the z score will likewise have “1-2-3” SD of mean.
T-Tests: used when population SD is not given
Hypothesis Testing: Z Scores One sample: when population is being
Sample statistics estimate population parameters with
compared against (given population mean)
varying degrees of success. Sampling error is the Related: When it’s the same person doing
difference between a statistic and its parameter. different tests (have more power than ind.
Standard Error is sampling error that tells u