Indiv Confidence Interval for more than 1 future data point

[#1]

In the Fit Y by X platform I oftentimes fit a simple linear regression model and choose to visualize the Indiv Confidence Interval around the fitted line. If I'm not mistaken, in essence this interval represents a Prediction Interval which defines where future individual datapoints will be with a given amount of certainty.

I assume the interval is calculated by default for just 1 future individual datapoint. But I'm not sure, also because the help function seems to mention multiple values:

Indiv Confidence Interval

Creates two new columns called Lower 95% Indiv <colname> and Upper 95% Indiv <colname>. These columns contain lower and upper 95% confidence limits for individual response values.

Is there a way to calculate/visualize this interval for an amount of future datapoints that can be specified? Or am I missing something and would it make no sense?

I can also refer to the calculation of a Prediction Interval in the univariate space, which can be accomplished in the Distribution platform by choosing the 'Prediction Interval' option. Here, one can choose the number of future individual datapoints upfront.

Thanks

Guy

[#2 · Accepted solution]

Hi,

The individual confidence intervals in the Fit-Y-by-X platform are used to estimate where 95% (for example) of the individual points will appear for a given value of X. If your x-axis is time, you could use it to project into the future, but the further you look, the less certain the prediction becomes.  Please take a look at the Time Series Forecast platform in JMP 15 (https://www.jmp.com/support/help/en/15.1/?os=win&source=application&utm_source=helpmenu&utm_medium=a...).  You may find that more appropriate for your application.  Let me know if this helps to answer your question.

[#3]

Thanks Hadley. According to your definition of the interval it sounds like I should rather consider it as a "sample interval" (an interval that captures a given proportion (eg 95%) of a population) rather than an actual prediction interval (that uses 95% as a measure of certainty).