biostats last minute

inductive reasoning

specific observations —> general results

deductive reasoning

general observations —> specific results

deductive vs. inductive reasoning

deductive: general observations —> specific results

inductive: specific observations —> general results

abductive reasoning

incomplete clues —> most plausible explanation

provide an interpretation for a 95% confidence interval caculated

in roughly 95% of samples from the population, when we compute 95% CI, the interval will include the true population mean

correlation between standard error and sample size?

larger sample size —> smaller standard error

why does larger sample size result in smaller standard error?

larger samples provide a more reliable estimate of the true population mean

how do you calculate standard deviation from a frequency table?

n = sum of the frequencies

add up all x*f

divide x*f by the frequency total (n)

this gives you the average

compute s as you normally would, multiplying each (x-avg)² by corresponding frequency

what assumptions are you making when you calculate the standard error of the mean of the sample?

the sample is random from the total population

what is significance level? alpha

pre-determined threshold that represents maximum probability of committing Type I error

when do you use the Bonferroni correction? what is it? what’s a downfall?

when you perform multiple hypothesis tests on the same data at the same time

divide alpha (significance level) by the total number of tests or comparisons

controls Family-Wise Error Rate —> probability of one False Positive

increases type II error, reduces power

reducing significance level does what to probability of committing Type I error?

decreases

how does the p-value relate to significance level when rejecting the null hypothesis?

• If p-value < a, you reject the null hypothesis.

• If p-value > a, you fail to reject the null hypothesis.

reducing significance level does what to probability of Type II error?

increases

reducing the significance level does what to the Power of a test?

decreases power

what impact does reducing significance level have to the sample size?

no effect

what does increasing sample size have on type 1 error? type 2 error?

no effect on type 1 error

decreases type 2 error

what effect does increasing sample size have on power and significance level?

increases power

no effect on the significance level

how do you calculate power?

power = 1 - B

B is probability of making Type II error

ways to reduce bias

control

randomization

blinding

ways to limit error

replication

balanced design

blocking

matched-pair design

extreme treatments