A z score of 1.645 means a value is 1.645 standard deviations above the mean of a distribution. For a standard normal distribution, it is approximately the 95th percentile: about 95% of values fall below it and 5% fall above it.
What a z score measures
A z score expresses a value’s position relative to a distribution’s mean in units of standard deviation. It is calculated as:
z = (x − μ) / σ
Here, x is the observed value, μ is the population mean, and σ is the population standard deviation. A positive z score is above the mean; a negative one is below it. A score of zero is exactly at the mean.
For example, if test scores have a mean of 70 and a standard deviation of 10, a score of 86.45 has z = (86.45 − 70) / 10 = 1.645. This interpretation assumes those mean and standard deviation values are appropriate for the data being assessed.
What percentile is 1.645?
In the standard normal distribution, the cumulative probability to the left of z = 1.645 is about 0.95. Thus, the value is at approximately the 95th percentile. The area to its right is about 0.05, and the area between the mean (z = 0) and 1.645 is about 0.45.
Percentile language describes a position in a distribution, not a percentage score. A 95th-percentile value is not necessarily 95% correct, nor does it mean that the value is 95% larger than average. It means that roughly 95% of the distribution lies at or below that point, if the standard normal model applies.
The number is rounded. More precisely, the z value with a cumulative probability of 0.95 is about 1.64485, commonly reported as 1.645. Small differences in tables or software can result from rounding.
Why 1.645 is used as a critical value
The value 1.645 is especially familiar as a critical value for a one-sided test at a 5% significance level under a standard normal model. A one-sided test asks whether a parameter is greater than a specified value or less than it, in a direction chosen before examining the results.
For an upper-tail test, a test statistic above 1.645 falls in the uppermost 5% of the standard normal distribution. For a lower-tail test, the corresponding cutoff is −1.645. The sign changes because the rejection region is on the opposite side of the distribution.
For a two-sided test with a total 5% significance level, the 5% is split between both tails: 2.5% in each. The standard normal critical values are approximately −1.96 and +1.96, not ±1.645. Confusing one-sided and two-sided cutoffs is a common source of errors.
Connection to confidence intervals
For a normally distributed statistic with a known standard error, a one-sided 95% confidence bound uses 1.645 as its multiplier. For example, an upper one-sided bound is often calculated as an estimate plus 1.645 times its standard error; a lower bound subtracts that amount. The exact setup depends on the question and the statistical model.
By contrast, a conventional two-sided 95% confidence interval under the standard normal approximation uses about 1.96 standard errors on either side of the estimate. That interval captures the middle 95% of the normal distribution, leaving 2.5% in each tail.
These normal critical values are not automatically suitable for every sample. When estimating a mean with an unknown population standard deviation, procedures commonly use a Student’s t distribution, especially for smaller samples. Its critical value depends on degrees of freedom and may differ from 1.645.
How to use the value correctly
Before applying 1.645, identify the distribution and the probability convention. A practical checklist is:
- Check standardization: Is the statistic expressed in standard deviation or standard error units, so that a z-based comparison is meaningful?
- Identify the tail: Is the question asking about an upper tail, a lower tail, or both tails?
- Confirm the probability: For 1.645, the cumulative probability to the left is about 0.95; the upper-tail probability is about 0.05.
- Check the model: Is a normal approximation justified, or should a t distribution or another method be used?
- Interpret the result in context: A threshold comparison alone does not measure practical importance or prove a claim is true.
A z score can also be converted back to the original measurement scale using x = μ + zσ. If the mean is 100 and the standard deviation is 15, z = 1.645 corresponds to 100 + (1.645 × 15) = 124.675, or about 124.7. This conversion relies on using the same mean and standard deviation that define the z score.
Common misunderstandings and limitations
A z score does not by itself establish that an observation is unusual in every setting. The percentile interpretation of 1.645 assumes a standard normal distribution, or a statistic that is adequately approximated by one. If the underlying distribution is strongly skewed, has heavy tails, or contains outliers, translating a z score directly into a normal percentile can be misleading.
Also, a value above a critical threshold is not proof that a research hypothesis is correct. In significance testing, the threshold helps define a decision rule under specified assumptions. The p-value, effect size, study design, uncertainty, and consequences of errors all matter. Statistical significance and practical importance are different questions.
Finally, distinguish standard deviation from standard error. A z score for an individual observation is often based on a standard deviation. A z statistic for an estimate, such as a sample mean, is typically measured in standard errors. Both use standardization, but they answer different questions.
Frequently Asked Questions
Is a z score of 1.645 good?
A z score describes relative position, not quality. Whether it is good depends on what is measured: it may indicate a high result on a test, or an unusually high value for a measurement where lower is preferable.
Is 1.645 the 95th percentile?
Approximately, yes, for a standard normal distribution. Its cumulative probability is about 0.95, so about 95% of values lie below it.
Is 1.645 used for a 95% confidence interval?
It is used for a one-sided 95% confidence bound under a standard normal approximation. A two-sided 95% confidence interval typically uses about 1.96, not 1.645.
What is the z score for a 5% significance level?
For a one-sided test, the standard normal cutoff is about 1.645 in the upper tail or −1.645 in the lower tail. For a two-sided 5% test, the cutoffs are about ±1.96.
How do I find the value that corresponds to z = 1.645?
Use x = μ + 1.645σ, where μ is the relevant mean and σ is the standard deviation. For an estimate-based test, use the appropriate standard error and procedure instead.
