Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

- JMP User Community
- :
- Discussions
- :
- Regression Analysis

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

May 20, 2017 3:29 AM
(956 views)

Hello Everyne.

One new question: I have a regression line (please see attachment) Y xs X.

I have calculated the confidence intervals for Y as a function or X.

However based on this fit, when I want to get the confidence intervals of X how do I calculate them.

The reason for this question is that ofter times we obsere a response (Y) and then read back the X value from the regression line. Therefore when an X value is read out of this regression line, it should theoretically also have a confidence interval ?

I am much obliged for your kind assistance.

1 ACCEPTED SOLUTION

Accepted Solutions

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

May 20, 2017 4:57 AM
(1875 views)

Solution

This is known as *inverse prediction* or the *calibration problem*. You have a standard curve and correctly fit Y (response) versus X (known levels). This result is available but not in the Bivariate platform that you used. Use the Fit Least Squares platform instead. Select **Analyze** > **Fit Model**. Enter your Y column in the Y role, enter your X column as an Effect, and click **Run**. Click the red triangle at the top and select **Estimates** > **Inverse Prediction**. Choose your **confidence level** and your desired **type of interval**. Enter up to **eight Y values**. Finally, use the **option** at the bottom to indicate if you want an interval estimate for the mean X (default) or the individual X value (check the box).

Learn it once, use it forever!

5 REPLIES

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

May 20, 2017 4:50 AM
(946 views)

~~You can get the confidence interval for your X variable by running the Distribution Platform.~~

See Mark's response below

Jim

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

May 20, 2017 4:58 AM
(942 views)

Hi there,

You mean distribution X as Y (right) ?

But this is not very helpful, it generates no meaningful report. I would be obliged for your assistance .

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

May 20, 2017 4:57 AM
(1876 views)

This is known as *inverse prediction* or the *calibration problem*. You have a standard curve and correctly fit Y (response) versus X (known levels). This result is available but not in the Bivariate platform that you used. Use the Fit Least Squares platform instead. Select **Analyze** > **Fit Model**. Enter your Y column in the Y role, enter your X column as an Effect, and click **Run**. Click the red triangle at the top and select **Estimates** > **Inverse Prediction**. Choose your **confidence level** and your desired **type of interval**. Enter up to **eight Y values**. Finally, use the **option** at the bottom to indicate if you want an interval estimate for the mean X (default) or the individual X value (check the box).

Learn it once, use it forever!

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

May 20, 2017 5:09 AM
(936 views)

This result is available but not in the Bivariate platform that you used.

**Use the Fit Least Squares platform** instead.

Sorry, could not find this platform ?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

May 20, 2017 5:24 AM
(933 views)

The platform that Mark is referring to is:

Analyze==>Fit Model

Within it you can choose the Least Squares

Jim