topic Re: How to calculate f score with 95% confidence interval using Base SAS in SAS Programming

How to calculate f score with 95% confidence interval using Base SAS

DeepakSwain — Thu, 21 Jun 2018 18:18:56 GMT

Hi there,

I am having a 2X2 table. I want to calculate f score with 95% confidence interval. Can anybody suggest me how to proceed.

/*CREATING TABLE 2 BY 2 TABLE */
data study;
input obs1 obs2 count;
datalines;
1 1 331
1 0 13
0 1 2
0 0 654
;
run;

Thank you in advance for your kind reply.

Regards,

Re: How to calculate f score with 95% confidence interval using Base SAS

PaigeMiller — Thu, 21 Jun 2018 18:49:44 GMT

You are talking about PROC FREQ, or something else?

As far as I know, there are no F-scores in a 2x2 table, there is a Chi-square value (not really a score), and the Chi-square value doesn't have confidence intervals associated with it.

So it's really not clear what you want. Please explain further.

Re: How to calculate f score with 95% confidence interval using Base SAS

DeepakSwain — Thu, 21 Jun 2018 18:55:55 GMT

Hi PaigeMiller,

Thank you for the reply.

I mean to say F1 score which is the harmonic mean of precision and sensitivity (i.e. 2TP/(2TP+FP+FN).

TP=true positive, FP=false positive, TN=true negative, FN=false negative

With regards,

Re: How to calculate f score with 95% confidence interval using Base SAS

PaigeMiller — Thu, 21 Jun 2018 19:02:49 GMT

If this is what you mean

https://en.wikipedia.org/wiki/F1_score

then I believe you could run PROC FREQ and then apply the formula for F1 to the PROC FREQ output. I don't see any confidence interval calculations.

Re: How to calculate f score with 95% confidence interval using Base SAS

DeepakSwain — Thu, 21 Jun 2018 19:51:25 GMT

Hi Paige Miller,

Just like kappa statistics with 95% CI can be obtained using proc freq with AGREE. I was looking for F1 score. Can you suggest something.

With regards,

Re: How to calculate f score with 95% confidence interval using Base SAS

PaigeMiller — Thu, 21 Jun 2018 19:55:52 GMT

I am not aware of a 95% CI for the F1 score, nor is Wikipedia aware of such a confidence interval.

Thus the only possibility I can think of is some sort of bootstrap.