Confidence interval for the median - Statistica General Discussion - Statistica - Dell Community

# Confidence interval for the median

#### Confidence interval for the median

For most methods, Statistica calculates (and usually displays) the confidence interval associated with a result.

When it comes to single samples, it seems to be a bit more difficult to get this info.   For a single proportion, we can use "Power Analysis > Interval Estimation > One Proportion, Z, Chi-Square Test".   However, for a single non-normal continuous variable, is it possible to calculate the confidence interval for a median?

For example, if we are wanting to compare a non-normal continuous variable to a target/reference value, we can use this workaround ("How to Perform a Wilcoxon Signed-Rank Test for One Sample") to obtain a significance value... but how do we see the confidence intervals associated with that comparison?

• Hi, as far as I know, calculation of confidence intervals for a median is currently not available in Statistica. If you keen, here is a website which describe some methods to calculate the CI for a median in single sample for your reference.

The workaround of Wilcoxon Signed-Rank test for one sample, assigns signed ranks to the differences between the sample values and compares the ranks to the rank of the reference/target value. The significance test is performed based on the assumption of the unit normal distribution of the sum of the non-zero signed ranks in the sample. You can refer to this website which has a detailed description of how Wilcoxon signed-rank test is performed.  This signed rank is not real observed values and confidence intervals for the signed rank is of little real value meaning and usually not calculated. CI for signed rank is also not calculated in Statistica.

Thank you.

All Replies
• Hi, as far as I know, calculation of confidence intervals for a median is currently not available in Statistica. If you keen, here is a website which describe some methods to calculate the CI for a median in single sample for your reference.

The workaround of Wilcoxon Signed-Rank test for one sample, assigns signed ranks to the differences between the sample values and compares the ranks to the rank of the reference/target value. The significance test is performed based on the assumption of the unit normal distribution of the sum of the non-zero signed ranks in the sample. You can refer to this website which has a detailed description of how Wilcoxon signed-rank test is performed.  This signed rank is not real observed values and confidence intervals for the signed rank is of little real value meaning and usually not calculated. CI for signed rank is also not calculated in Statistica.

Thank you.

• Thanks so much Jenny, very informative.

• You are most welcome :)