Sample size to estimate Pearson's Correlation (P Coefficient (Part-1) | Scribe

    Sample size to estimate Pearson's Correlation (P Coefficient (Part-1)

    • Sudhir Verma |
    • 7 steps |
    • 38 seconds
      1

      Sometimes aim of clinical research is find whether there is correlation between two numerical variables. For example, we may be interested in knowing whether systolic blood pressure is correlated with body mass index (BMI). We design a study and now want to know how many participants we should enrol in the study.

      information ordinal icon

      We will need data from prior studies or pilot studies about expected correlation. Suppose, previous studies have reported correlation coefficient to be approximately 0.4. Sample size calculation can be done by two approaches, in the first approach we test null of hypothesis of no correlation (i.e. we compare expected r=0.4 with Null r=0) and in second approach we would like to estimate precision around correlation coefficient. Let us see how to do it

      2

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

      3

      Click "Pearson's correlation"

      4

      Click expected correlation field.

      5

      Enter expected correlation as 0.4, let us keep other fields as defaults i.e. significance level at 0.05, power at 80% and dropout rate at 10%. Hit the calculate button

      6

      This calculated sample size depicts number of participants required in order to test whether expected correlation coefficient is different from null correlation at significance level of 0.05 and power of 80%. Since we have assumed a larger effect i.e. difference in r of 0 and r of 0.4, we need lesser sample size.

      7

      Now if we use expected correlation coefficient 0.3 instead of 0.4, required sample size will be increased