I have a data set with a series of patient encounters where the patient went from a hospital into a nursing facility and the number of days they stayed in the nursing facility like this:
| Patient | Hospital | Nursing Facility | Days |
| 1 | Hospital A | Facility 1 | 20 |
| 2 | Hospital A | Facility 1 | 12 |
| 3 | Hospital B | Facility 1 | 23 |
| 4 | Hospital B | Facility 2 | 45 |
And it's rolled up like this:
| Hospital | Nursing Facility | Average Length of Stay |
| Hospital A | Facility 1 | 35 |
| Hospital A | Facility 2 | 32 |
| Hospital B | Facility 2 | 22 |
| Hospital C | Facility 3 | 25 |
| Hospital C | Facility 1 | 45 |
And I've brought in some comparison data as well where both the Hospital and Nursing Facility are labelled 'Comparison':
| Hospital | Nursing Facility | Average Length of Stay |
| Hospital A | Facility 1 | 35 |
| Hospital A | Facility 2 | 32 |
| Hospital B | Facility 2 | 22 |
| Hospital C | Facility 3 | 25 |
| Hospital C | Facility 1 | 45 |
| Comparison | Comparison | 24 |
What I want to do is add a column that gives me whether the current row is statistically significantly different than the comparison data with a t-test:
| Hospital | Nursing Facility | Average Length of Stay | p value of difference from comparison |
| Hospital A | Facility 1 | 35 | .002 |
| Hospital A | Facility 2 | 32 | .007 |
| Hospital B | Facility 2 | 22 | .2 |
| Hospital C | Facility 3 | 25 | .4 |
| Hospital C | Facility 1 | 45 | .000001 |
| Comparison | Comparison | 24 | - |
And I'm having trouble with the set analysis necessary to make this happen. TTest_sig requires a group field, and a value field. Those should be the NursingFacility and LegnthOfStay. It also requires that the group field have 2 and only 2 groups in it. What I want to do conceptually is something like this:
TTest_sig({<current row>+<comparison data>} NursingFacility,LengthOfStay)
But the set analysis syntax totally escapes me. Any help would be very much appreciated.
