EPI- Chapter 8

power

1-beta probability of correctly concluding that the treatments differ

o The treatments are not different and we conclude that they are not different (correct) o there is no difference in efficacy between treatments, but we conclude that there is a difference (Type I error) o there is a difference in efficacy between therapy A and therapy B, but we conclude find no difference between them (Type II error) o there is a difference in efficacy between therapy A and therapy B, and when we examine the groups in our study we find that they differ. (correct decision) (power)

4 possibilities in testing whether the treatments differ

"gold standard" because it is used to minimize or avoid selection and other types of biases

Advantages of randomized trials

Is randomization ethical? Is knowingly withholding treatment from a patient ethical? Generalizability-->Study population not always indicative of general population

Disadvantages of randomized trials

1-beta = power

How are beta and power related?

o must know to what extent the patients we have studied are representative of the defined population o must characterize those who did not participate in the study and identify characteristics of study patients that might differ from those in patients who did not participate in the study.

What is required before external validity can occur?

o Difference in response rates to be detected o Estimate of the response rate in one of the groups o Level of statistical significance (alpha) o Value of power desired (1-beta) o Whether test is one or two sided

What must be specified to estimate the sample size needed for a randomized trial?

external validity (generalizability)

ability to apply results obtained in a study population to a broader population (total population or defined population)

type II error

error in which you conclude that the treatments DO NOT differ when in reality they do

Type I error

error in which you conclude that the treatments differ when in reality they do not

p > alpha

fail to reject the null hypothesis, there is NO difference between the treatments

internal validity

how well an experiment is done, especially whether it avoids confounding was the research done right? is the trial randomized and free of bias?

alpha

probability of concluding that the treatments differ when in reality they do not

power

probability of detecting a difference between the treatments if they do in fact differ

alpha

probability of making a type I error

beta

probability of making a type II error

p-value

probability of obtaining a result equal to or "more extreme" than what was actually observed, when the null hypothesis is true

beta

probability of probability of concluding that the treatments DO NOT differ when in reality they do

p < alpha

reject the null hypothesis, there is a difference between the treatments