Continuity correction is the 0.5 adjustment you make when a continuous normal distribution is used to approximate a discrete distribution. In Intro to Probability, it tightens probabilities for binomial and Poisson problems.
Continuity correction is the 0.5 adjustment you use when a continuous distribution, usually the normal distribution, is approximating a discrete one in Intro to Probability. It accounts for the fact that discrete counts happen in whole numbers, while the normal curve spreads probability across an interval.
The basic idea is simple: a discrete value like 8 is not a single point on a normal curve, it stands in for the entire interval around 8. So when you want the probability of exactly 8, or of 8 or less, you shift the cutoff by 0.5 to catch the right chunk of area under the curve.
For example, if you are approximating P(X <= 8) for a binomial random variable with a normal distribution, you usually use P(Y <= 8.5). That extra half unit captures the whole mass at 8, since the normal model does not have separate bars for each integer. If you are finding P(X >= 8), you would usually move the boundary the other way and use 7.5.
This is why continuity correction shows up most often with binomial and Poisson approximations. Those distributions count events, so their values jump in integers, but the normal distribution is smooth. The correction makes the smooth curve line up more closely with the stepped shape of the discrete distribution.
A common mistake is to forget which side gets the 0.5 shift. The direction depends on the wording of the probability statement. "Less than or equal to" and "greater than or equal to" are not handled the same way, and "between" problems usually need both ends adjusted outward so the interval covers the full bars on the discrete scale.
Continuity correction matters because it changes a rough normal approximation into one that better matches the actual discrete probability question. In Intro to Probability, that means your answer is more faithful to the distribution you started with, especially when you are working with binomial or Poisson random variables.
Without the correction, the normal curve can miss part of the probability sitting on a boundary value. That may not seem like much, but when the sample size is not huge or when you are asking for a probability near the center or tail, the difference can be noticeable. The correction is a quick way to reduce that error.
It also trains you to think carefully about what a probability statement really means. A question like P(X < 10) is not the same as P(X <= 10), and the continuity correction forces you to translate the language into the right interval on the number line. That translation skill shows up again and again in problem sets.
If you are using the normal approximation to the binomial, continuity correction is part of setting up the problem correctly, not just polishing the final answer. It connects the discrete counting model to the smooth curve model in a way that makes the approximation much more usable.
Keep studying Intro to Probability Unit 9
Visual cheatsheet
view galleryNormal Distribution
Continuity correction shows up when you use the normal distribution to estimate probabilities for a discrete random variable. The normal curve is smooth, so the 0.5 shift helps you line it up with integer counts on the discrete scale. If you are reading a z-score problem, this is the adjustment that keeps the area calculation tied to the original count question.
Binomial Distribution
Binomial problems are one of the most common places to use continuity correction. A binomial variable counts successes, so its values are whole numbers, but a normal approximation treats them like a continuous spread. The correction helps you move from the binomial count to the right normal cutoff, especially for cumulative probabilities.
Poisson Distribution
Poisson random variables count events in a fixed interval, which makes them discrete just like binomial counts. When a normal approximation is allowed, continuity correction helps the curve match the jumpy Poisson bars more closely. This matters in event-count problems where you want a probability for at most, at least, or between specific counts.
Quantiles
Quantiles are cutoff values that split probability into areas, and continuity correction can change which cutoff you use when the original variable is discrete. If a question asks for the value where a certain percentage falls below it, you still have to think about whether that boundary represents an integer count or a continuous percentile. The correction helps you translate between those two setups.
A quiz or homework problem will usually give you a discrete model, then ask you to approximate a probability with the normal distribution. Your job is to spot whether the statement is "less than," "at most," "greater than," or "between," then shift the cutoff by 0.5 in the correct direction before finding the z-score.
For example, if a problem asks for P(X <= 12) for a binomial random variable, you would use 12.5 as the normal cutoff. If it asks for P(X >= 12), you would use 11.5. On a problem set, the most common mistake is skipping the correction and calculating the area at the whole number instead of the adjusted boundary.
You may also be asked to explain why the correction is needed. A good answer says that discrete probabilities live at integer values, while the normal curve spreads area over intervals, so the 0.5 shift better matches the discrete bar being approximated.
Normal approximation is the larger method of using a normal curve to estimate a discrete probability. Continuity correction is the small adjustment inside that method that shifts the cutoff by 0.5. So the approximation is the full strategy, and the continuity correction is one step that makes it more accurate.
Continuity correction is the 0.5 shift you use when a normal distribution stands in for a discrete one.
It matters because whole-number counts from binomial or Poisson models need interval-based cutoffs on a normal curve.
For "at most" or "less than or equal to" statements, the cutoff usually shifts upward by 0.5.
For "at least" or "greater than or equal to" statements, the cutoff usually shifts downward by 0.5.
The correction is not extra decoration, it changes which area under the curve matches the original probability question.
It is the 0.5 adjustment used when you approximate a discrete distribution with a continuous normal distribution. Because discrete values happen at whole numbers, the correction shifts the boundary so the normal curve captures the right probability area. You will usually see it with binomial and Poisson problems.
No, you use it when a continuous normal distribution is approximating a discrete distribution. If the problem is already using a discrete distribution directly, there is no need for the 0.5 shift. In normal probability questions where the variable is already continuous, continuity correction does not apply.
Look at the inequality. For "at most" or "less than or equal to," move the boundary up by 0.5. For "at least" or "greater than or equal to," move it down by 0.5. For an interval like 5 to 10, you usually adjust both ends outward.
Because both distributions count outcomes in integers, while the normal curve is smooth. The correction makes the normal area line up more closely with the bar or point probability you started with. It is especially useful when the count is near a boundary or when the sample size is not huge.