Medical statistics and Data Science: Clinical Trials

Sample size calculation and power analysis

You may watch a Youtube video so called "Biologist talks to statistician" (link)

One sample test for means

  1. Compute sample size
  2. Significance level (α):
    Desired power (1- β):
    Two-sided or one-sided test
    Expected mean of the sample:
    Expected standard deviation of the sample :
    Mean of the reference group:

  3. Compute power
  4. Significance level (α):
    Sample size:
    Two-sided or one-sided test
    Expected mean of the sample:
    Expected standard deviation for the sample :
    Mean of the reference group:

Paired test for means

  1. Compute sample size
  2. Significance level (α):
    Desired power (1- β):
    Two-sided or one-sided test
    Expected mean of the difference (after - before):
    Expected standard deviation of the difference :
    Null hypothesis for mean of the difference:

  3. Compute power
  4. Significance level (α):
    Sample size:
    Two-sided or one-sided test
    Expected mean of the difference (after - before):
    Expected standard deviation for the difference :
    Null hypothesis for mean of the difference:

Comparing two-sample means

  1. Compute sample size
  2. Significance level (α):
    Desired power (1- β):
    Two-sided or one-sided test
    Expected mean for the intervention group:
    Expected standard deviation of the mean for the intervention group :
    Expected mean for the control group:
    Expected standard deviation of the mean for the control group :

  3. Compute power
  4. Significance level (α):
    Sample size per group:
    Two-sided or one-sided test
    Expected mean of the intervention group:
    Expected standard deviation for the mean of the intervention group :
    Expected mean of the control group:
    Expected standard deviation for the mean of the control group :

Comparing two-sample proportions

  1. Compute sample size
  2. Significance level (α):
    Desired power (1- β):
    Two-sided or one-sided test
    Expected proportion for the intervention group:
    Expected proportion for the control group:

  3. Compute power
  4. Significance level (α):
    Sample size per group:
    Two-sided or one-sided test
    Expected proportion for the intervention group:
    Expected proportion for the control group:

Two-sample hazard ratio using the Cox proportional hazards model

  1. Compute sample size
  2. Significance level (α):
    Desired power (1- β):
    Two-sided or one-sided test
    Expected hazard ratio by comparing the intervention group to the control group:
    Expected failure proportion ("number of cases"/"total number of sample size" at the end of the follow-up :
    Expected proportion of withdraw from the study:

  3. Compute power
  4. Significance level (α):
    sample size per group:
    Two-sided or one-sided test
    Expected hazard ratio by comparing the intervention group to the control group:
    Expected failure proportion ("number of cases"/"total number of sample size" at the end of the follow-up :
    Expected proportion of withdraw from the study:

Prevalence Studies

  1. Compute sample size
  2. Significance level (α):
    Expected prevalence among the population:
    Maximum acceptable error for the expected prevalence

  3. Compute maximum error
  4. Significance level (α):
    Expected prevalence among the desired population:
    sample size: