Standard Error for Difference in Proportions
What you'll learn: How to calculate the standard error when comparing two sample proportions using the pooled proportion.
When testing whether two populations have different proportions (like comparing conversion rates between two websites), we need to measure how much variability exists in the difference between our sample proportions. This is where the standard error comes in.
Why We Need a Pooled Approach
Under the null hypothesis, we assume both groups come from populations with the same true proportion. If they're equal, it makes sense to combine (or "pool") the data from both samples to get our best estimate of that common proportion. Think of it like mixing two buckets of water that should have the same temperature—you get a better reading by measuring the combined pool.
The Formula
The standard error for the difference in proportions is:
SE = sqrt(p_pooled × (1 - p_pooled) × (1/n₁ + 1/n₂))
Where:
p_pooledis the pooled proportion (total successes divided by total observations from both samples)n₁andn₂are the sample sizes of the two groups
Why This Form?
The (1/n₁ + 1/n₂) part accounts for sampling variability from both samples—smaller samples contribute more uncertainty. The p_pooled × (1 - p_pooled) part captures the variance of a proportion (larger when p is near 0.5, smaller near 0 or 1).
Key Takeaway: The standard error for difference in proportions uses the pooled proportion under the null hypothesis assumption of equality, and adjusts for both sample sizes to measure the expected variability in our observed difference.