Point Estimation - I need to use the Quadratic Regression Model - Minitab 14

[jazzyd]

Hi Guys,

I'm really hoping someone out there can help me.
I have been doing regression analysis in Minitab 14 and i need to do a point prediction. However, when i do a point prediction at the moment it always uses the formula for the Linear Regression model but i need to use the Quadratic regression model. Can anyone help?

[Tim Folkerts]

Welcome Jazzyd!

I'm not sitting at the right computer to explore the options within mintab, but would it work to get the answer "by hand"? If you only need to do this a few times, then a manual calculation (or perhaps using Excel) would also work.

Once you have the equation in the form
y = ax^2 + bx + c
and Minitab tells you the values of a, b, and c from the regression, you could plug in a value for x and find y. Or you could create a formula within Excel (or within Minitab for that matter) to calculate y for many values of x.

Of course, if you need to do this many times, or you need to plot the quadratic regression within Mintab, then calculating it yourself would be less useful.

Perhaps later I can get to the "right" computer and see about minitabs actual options.


Tim F

[jazzyd]

Hi Tim,

Thanks for the reply, basically my issue is that i am doing a report and within the report i have to submit proof that i have used Minitab, i.e. a printout from it which includes prediction and confidence intervals.
i've managed to get a prediction point with the linear regression line, but i can't see any way of doing it with the quadratic regression formula. I've looked through all sorts of help files and get nowhere. If you can get to the 'right' computer and help me it would be much appreciated

[Tim Folkerts]

Would a graph suffice as "proof"?

In version 14 of Minitab, look in the menu Stat/Regression/Fitted Line Plot. Click the button for "Options" and there are check boxes for "Display Confidence Interval" and "Display Prediction Interval". If you select those, then at least the prediction & confidence intervals will be displayed on the graph.

You could cut & paste the resulting graph into your report. I don't see a way to get the actual numerical values, but often a graph is as good as a bunch of number in a table.

Tim F

[reynald]

Originally Posted by Tim Folkerts

Welcome Jazzyd!

I'm not sitting at the right computer to explore the options within mintab, but would it work to get the answer "by hand"? If you only need to do this a few times, then a manual calculation (or perhaps using Excel) would also work.
Tim F

Jazzyd,
Fortunately for you, i am now using minitab.
Do this:
Double click the line in the scatter plot with linear fit. Then a dilog box "edit Regression fit" will pop-up.
Select Options tab, then click your options.

Hope that helps

>>rey

[Statistical Steven]

Jazzyd

Create a new variable which is x^2 then fit the model Y=x + x^2 which will give you the coefficients for each parameter. Let me know if this works.

[jazzyd]

Hi guys,

Thank you all for your help. I ended up using Steven's suggestion. Unfortunately Tim a graph was not sufficient evidence. Rey i tried your method on some old date for which i knew what the correct estimate was and it through out a whole load of stuff that i didn't really know how to comprehend. Thanks once again for your help guys.

Jaz