gradient: for every extra 1 in x,
you get an extra m in y
intercept: when x is zero,
y is expected to be c
Extrapolation
Summary Statistics
Standard Deviation
Variance
Mean
Estimate from Grouped Data
Median and Quantiles
n+1 for individual items,
just n for grouped data
Skew
Formal definition
Annotations:
for positive:
o (Q2-Q1) < (Q3-Q2), i.e. right side of box fatter
o OR mean > median > mode, i.e. the lump of the mode happens early on, then there's more spread on the positive side
o 3(mean - median) / sd
Outliers: 1.5xIQR away from quartiles
Interpolation
Annotations:
it's basically about how far through the relevant class our wanted figure is.:
(how many items into this group) / (frequency of this group) x class width + class lower bound
When Comparing Data,
USE YOUR ASS!
Annotations:
Comment on:
o averages (whichever is handy)
o spread (sd, iqr, range)
o skew
use proper statistical names for things, not just "average" or "spread"
Graphs and Charts
Stem and leaf
Histogram
Cumulative Frequency
Curve for continuous, "steps" for discrete
Box plots
Probability
Discrete Random Variables
Expectation and Variation
Expectation Algebra,
Var(aX+b), E(aX+b)
Cumulative Probability F(x)
simultaneous equations,
use sum of p = 1
Normal distribution
Two sketches every time
Mapping ("bridge") equation: z=(x-mu)/sd
if mean and sd missing, set
up two bridge equations
if you have a prob. or a
%, do reverse-lookup in
tables to get z value
Trees
Venn Diagrams
Work from the
intersections
outwards, P(A) means the whole circle