There are no methodological method I know to do this without any data that would indicate what a 'No' would look like.
You could do distance measures between the observations and then pick any that are far away in terms of distance, though as I type this, I think that's likely mathematically the same as Principal component analysis. The next step would be to do this for a bunch of different observations and make decisions and see how it works and making cutoff rules but you have no data to do that.
What I would start off with is by taking the two samples of data and seeing if they're different across all the variables. Then I would probably start with those variables as my method of differentiating the groups.
This is an example of how that would work and would run a t-test on all the numeric variables across the data to see which are different. If the values are not different between the data sets they likely cannot differentiate between Yes and No because it's all the same as the Yes datasets. This could be misleading if you have a small event rate for No.
data have;
set GroupA GroupAB indsname = source;
dsn = source;
run;
proc ttest data=have;
class dsn;
var _numeric_;
run;
@karaheller20 wrote:
SO, the data I have currently is just with the data that is a yes. I have a few variables to predict it off of, but I need to come up with a way to predict whether it is a yes or a no based off the dataset that is just a yes. I have another dataset with the same variables but do not know if they are yes or no, so ideally I would like to apply the model I find in the dataset that is only yes, to this data set to determine if they are yes or no. Please let me know if this does not make sense!
Thoughts, in the past I have run detailed logistic regression but it was on a dataset that had both yes and no. This has me thinking that I cannot use logistic regression for this because the dataset I have to use to determine the results is just a yes.
I am using SAS enterprise guide
... View more
