Stats terminology and tests Flashcards
What must be disproved in order to accept the alternate hypothesis?
the null hypothesis
General example of null hypothesis?
x has no effect on y
General example of an alternate hypothesis?
x does have an effect on y
When are two samples said to be significantly different?
when p<0.05
What is the p value generated from a t test?
the probability that two samples have come from the same population by chance
What type error is shown in the p value?
The odds of rejecting the null when the null is true
What is a type 1 error?
Rejecting H0 when it is true
What do you conclude under a type 1 error?
That a difference exists when one does not exist
What is a type 2 error?
Accepting H0 when it is actually false
What do you conclude under a type 2 error?
No difference when one does exist
What advantages do parametric tests have over non-parametric tests?
More resilient to violations of assumptions and are more powerful
What is skew in data?
When the apex of the curve is to the left or the right
Positive skew?
Apex of curve is to the left
Negative skew?
Apex of the curve is to the right
Zero skew?
Apex of curve is in the middle
What is kurtosis?
The size and height of the curve
Platykotic?
Curve is wider (drops off later)
Mesokurtic?
Normal distribution, no unusual width of curve or height
What is leptokurtic?
Curve is a lot thinner and apex is higher
What is the null hypothesis for tests of normality?
That the distribution is normal
Tests for normality?
Shapiro-wilk, Kolmogorov-smirnov, Anderson-darling
How can you convert non-normal data into normal data?
By transforming it
Examples of a possible transformation?
Log transformation, square root, arcsine
What are log transformations used for?
positively skewed data
What are square root transformations used for?
Data that follows a poisson distribution
What are arcsine transformations used for?
Data using proportions or
percentage
What does a t test compare
the means of two samples
What does a t test estimate?
the probability that the observed means have come from the same population by chance
What are the 3 types of t test?
One sample, independent, and paired
What assumptions are used when doing a t test?
Normally distributed data
One sample t test null hypothesis?
The mean of a sample group is not significantly different from the overall mean
One sample t test alternate hypothesis?
The mean of a sample group is significantly different from the overall mean
Question that you would try to answer with a one sample t test?
Does the observed mean of x sample differ enough from the mean of y population to suggest that there may be something causing it, or could it have occurred by chance
When do you do a two tailed t test?
When you don’t have an idea of the direction of the difference. (e.g. you have no idea what effect z variable will have on x sample to make it different from y mean)
When do you do a one tailed t test?
When you do have an idea of the direction of the difference (e.g. you think that z variable will have w effect (making it higher or lower) on x sample to make it different from y mean)
Paired t test null hypothesis?
the true mean difference between the paired samples is zero (i.g. there isn’t any difference between the values before and after the variable has been applied)
If the null hypothesis in a paired t test was true, what would all observable explained differences be explained by?
random variation
Paired t test alternate hypothesis?
The true mean difference between the paired samples is not equal to zero (there is a difference between the values before and after the variable has been applied)
If the null hypothesis in a paired t test was rejected what could the observable differences be explained by?
The variable that has affected them (e.g. conditions grown in, training done etc)