Please enable JavaScript.
Coggle requires JavaScript to display documents.
Ch 2: probability models (bootstrap samples: the large number of samples…
Ch 2: probability models
bootstrap samples: the large number of samples drawn after an initial sample has been taken from the population
caluculate sample statistic for each bootstrap sample and collect them as sampling distribution (approx. 5,000 samples needed)
-
sampling distribution represented in histogram-- shows prob of every proportion / # of sample statistic in the sample
check is b. sample represents true sample distribution by comparing proportion of s. statistic in OG sample to proportion from bootstrap samples
if we bootstrap without replacement, our bootstrap sample will be identical to the initial sample
-
-
-
-
-
-
-
with large samples, the sampling distribution becomes more peaked bc there is a lower spread
-
-
-
rule of thumb: you can use normal distribution as approximation of sampling distribution if prop of s. statistic * sample size= greater than 5
-
a large sample is needed when a distribution is skewed/not symmetrical (further away from 0.5)--> large sample needed to correct
-
-
-
-
-
3) interpret results-- levene's test= not sig?--> check first row in "bootstrap for ind. s. t-test"-- "b.strap" table--> mean diff.-- CI
-
interpreting: strong/weak association (Cramers V) btwn (insert variables), significance and p-value, final statement (what does it mean)
-
-
create sampling distributions through: exact, bootstrapping and theoretical probability distribution