Sample Size for Two Independent Means | Scribe

    Sample Size for Two Independent Means

    • Abhijit Pakhare |
    • 13 steps |
    • 2 minutes
      information ordinal icon

      Let us learn how to calculate sample size for estimating difference between two means. Suppose as researchers we wanted to test whether mean systolic blood pressure is lower among those treated under combination therapy as compared to mono therapy. We have planned a RCT and now want to estimate number of patients who need to be enrolled in the trial per arm.

      information ordinal icon

      We will need some prior data or data from pilot study which will form the basis of sample size. We have searched the literature and found that, there is difference of around 5 mmHg between two treatment modalities. We will also need mean values and standard deviation in each group to calculate pooled standard deviation. From these studies we have computed pooled standard deviation as 15 mmHg.

      alert ordinal icon

      We need to take few decisions before proceeding for sample size calculations. We have to fix Type-I error and Type-II error which indicates level of confidence and power of the study. Let us set them as 0.05 (95% confidence level) and 0.2 (80% power) respectively. Now we are ready to compute sample size.

      1

      Navigate to [https://wnarifin.github.io/ssc_web.html](https://wnarifin.github.io/ssc_web.html)

      2

      Click "Two-mean comparison (independent)"

      3

      Click standard deviation field.

      4

      Type "15"

      5

      Click on expected difference number field.

      6

      Type "5"

      7

      Click on significance level or alpha field and keep it as default for now. Also leave power and expected dropout rate to default.

      8

      Click on calculate and you will get sample size.

      alert ordinal icon

      Remember this sample size is per group and total required sample size will be double of this number.

      9

      Now, let us explore by changing our assumptions i.e. decrease the expected difference

      10

      If expected difference is lesser we will require more sample size to test the hypothesis.

      11

      Now, let us explore by changing our assumptions i.e. increase the power of study to 90%

      12

      We can see that if we want more power for our study results then we need to enrol more participants.

      information ordinal icon

      How one should write about sample size calculation in the protocol. "Sample size calculations were done to by using a web based sample size calculator. (Arifin, W. N. (2022). Sample size calculator (web). Retrieved from http://wnarifin.github.io) Calculations are done by setting Type-I error at 0.05 and power at 80% with expected difference of approximately 5 mmHg and standard deviation of around 15 mmHg. Sample size required per arm of the study is 142 participants and after adjusting for drop out rate of 10%, it will be 158 participants. Thus a total of 316 participants will be randomised to either arm."

      13

      external-screenshot