New Contributor
Posts: 4

# type 1 error

I work in multivarite (high dimension) and use robust estimators for that I want to calculate type 1 error

how can I do that by using SAS I found many articles use SAS but no one explaine how

best regards

Super User
Posts: 10,213

## Re: type 1 error

Just a guess. Wouldn't it just be ALPHA value. SAS default is 0.05 ?

New Contributor
Posts: 4

## Re: type 1 error

compare the robustness (Type I error)

Posts: 2,655

## Re: type 1 error

Robustness and type I error usually have different meanings statistically, although they are connected.

Robustness says something about how much departures from the assumptions of the analysis affect point and interval estimates, and resulting tests. (Yes, it is more complicated than that, but for now that is a good enough definition).  Type I error is the probability that a deviation as large as the one observed could originate due to chance alone, provided that both the null hypothesis is true AND the assumptions of the analysis are met.

So, let's say that for a particular analysis, some assumption is not met--say that the residuals in a regression aren't normally distributed.  Then, the normal least squares regression is not robust, and type I error will be affected.  You can't say how much it is affected, as that will depend on how badly the assumption of normal errors is violated.

Does that help any?

Steve Denham

New Contributor
Posts: 4

## Re: type 1 error

You are true

in testing hypothesis, the part of robustness is ability of the procedures to control Type I error rate of a test to close to the significance level α

Posts: 2,655

## Re: type 1 error

You said:

"in testing hypothesis, the part of robustness is ability of the procedures to control Type I error rate of a test to close to the significance level α"

That's correct.  So what is question you are asking?  Is it about the robustness of some particular statistical method?  If so, what method?

The only way I know of to measure robustness is through simulation--and simulating enough datasets that violate an assumption to adequately test is difficult.

Steve Denham

Frequent Contributor
Posts: 140