setting tolerance criterion

dasher · **Posted:** Wed Jan 14, 2009 1:13 pm

Hi,

Is it possible to set a value for the tolerance criterion when constructing a multiple linear regression model? i.e. I tick the Collinearity diagnostics box to give me VIF and tolerance, but I've read that SPSS automatically sets the tolerance to 0.0001, that way too low for me but I cant see anywhere how to set a higher value.

Many Thanks in advance

Fierce · **Posted:** Wed Jan 14, 2009 1:53 pm

When you define Dependant, Factors, etc... You can define this in the "ESTIMATION" area.
You can also generate a Syntax for your model and change the number in "SINGULAR(<NUMBER HERE>)".

dasher · **Joined:** Wed Jan 14, 2009 5:56 pm **Posts:** 3

Hi Fierce,

I'm using SPSS 11.5 and I am using the options 'Analyse', 'Regression', 'Linear Regression'. When I get to the stage where I'm entering my dependents/independents, I cant see the 'estimates' box I have to tick..... :cry:

Fierce · **Posted:** Wed Jan 14, 2009 6:40 pm

Oh, sorry. This option is visible for Mixed models. My bad.

Anyway, you'll have to use syntax. Look for a the "/CRITERIA" in your syntax and modify it like in the example::

Code:

/CRITERIA=PIN(???) POUT(???) TOLERANCE (<Tolerance value here>)

I can't be sure if it will work in version 11,5 as I never used it (and I don't have anything from version 11,5). But it probably will. If it doesn't, please let me know.

dasher · **Joined:** Wed Jan 14, 2009 5:56 pm **Posts:** 3

Hi Fierce,

Many thanks for that. I'll get a friend to do it for me as I'm not familiar with entering code etc.

But am I correct in saying that SPSS has automatically set a criterion of 0.0001? If i know this for definate, this should do for the moment.

Many Thanks

D.

Fierce · **Posted:** Wed Jan 14, 2009 10:08 pm

Yes, the default tolerance value is 0.0001.

Ah, you don't need to know a lot of syntax code if all you have to do is modify a line. Just do it all with the menus and hit PASTE instead of OK, it will generate the syntax for you.

dasher · **Joined:** Wed Jan 14, 2009 5:56 pm **Posts:** 3

Hi Fierce,

Just one more question on tolerance criterion!

If SPSS has automatically set it to 0.0001, and I want to carry out a stepwise multiple linear regression procedure, SPSS does not use the tolerance value in any way to decide whether a variable stays in the model or not i.e. its just using a significant change in the F-statistic as the criterion for selecting/rejecting a variable.

Is the point of showing the collinearity statistics so that we use our own judgement?

Thankyou,

dasher

setting tolerance criterion

Who is online