statistics Flashcards
(32 cards)
dependent
value depends on the other value
drawing scatter graphs
if applicable, controlled x, dependent y
Interpolation
Between upper and lower bounds of data
Extrapolation
Beyond given data (less reliable, relationship may not continue to be linear)
Steep + large positive syy
small sxx
shallow + positive/negative
small syy, large sxx
y on x
y=a+bx
predicting y given x
minimise distances in x-direction
residuals
distance of a point from a regression line
Σresiduals
0
r
-1≤ r ≤1
measure of correlation
PMCC conditions
data evenly spread within an ellipse
bivariate normal distribution (2 variables, continuous, random)
correlation
How close data is to a line
PMCC
Testing for correlation in the population sample
1-tailed
Specific correlation
2-tailed
Any correlation
first statemrnt pmcc
Let p (rho) be the population correlation coefficient
null hypothesus PMCC
There is no correlation between x and y in a population
If defined, p = probability
coefficient of determinant r^2
Proportion of the variation in y that is explained by the variation in x
How much h changes as x changes
Higher r^2 closer to line
regression line random on non-random
y on x
pmcc reject/accept
|r| > c.v reject H0
residual
Observed - expected
Effect sizes over PMCC
Large set of random bivariate data
Small non-zero PMCC, reject H0
Uninformative
Better model requires
Not always suitable for large sample because it has a low critical value
cohen s,m,l
0.1,0.3,0.5
5% significance
Probability of getting large PMCC where there is no PMCC in population is 5%
5% chance of incorrectly rejecting H0