STATA v15.1
I have 3 level data.. patients, surgeons, and facilities
Surgeons and facilities are crossed, as surgeons can operate at more than one facility and I also included a random interaction that was significant
I'm trying to parse the variance explained at each level, including how much variance is explained by my patient-level fixed effects
---
My naive model (no pt lvl effects) is...
mixed caseprice || _all: R.loc || surgeon: || loc:
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
_all: Identity |
var(R.location) | 18382.64 14440.07 3942.441 85713.72
-----------------------------+------------------------------------------------
surgeon: Identity |
var(_cons) | 7339.356 4842.625 2013.824 26748.19
-----------------------------+------------------------------------------------
location: Identity |
var(_cons) | 5477.803 3445.984 1596.362 18796.7
-----------------------------+------------------------------------------------
var(Residual) | 53149.89 1783.819 49766.18 56763.67
----
Then with all pt level fixed effects (including multiple imputation)...
mi estimate: mixed caseprice c.age_dm##c.age_dm i.race_short bmi i.asa_score i.asa_emergent i.urgency c.elixsum##c.elixsum i.indication i.resident_yn || _all: R.loc || surgeon: || loc:
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
_all: Identity |
sd(R.location) | 99.45665 43.09123 42.54385 232.5042
-----------------------------+------------------------------------------------
surgeon: Identity |
sd(_cons) | 98.62659 27.50079 57.10145 170.3495
-----------------------------+------------------------------------------------
location: Identity |
sd(_cons) | 70.78175 22.57506 37.88261 132.2521
-----------------------------+------------------------------------------------
sd(Residual) | 223.3151 3.74665 216.0912 230.7805
With these results, I believe, my ICC surgeon is ~ 13%, ICC location ~ 13%, ICC for combination of surgeon and location is 33%, but not sure what I can (if anything) say about the patient level variance explained?
I have 3 level data.. patients, surgeons, and facilities
Surgeons and facilities are crossed, as surgeons can operate at more than one facility and I also included a random interaction that was significant
I'm trying to parse the variance explained at each level, including how much variance is explained by my patient-level fixed effects
---
My naive model (no pt lvl effects) is...
mixed caseprice || _all: R.loc || surgeon: || loc:
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
_all: Identity |
var(R.location) | 18382.64 14440.07 3942.441 85713.72
-----------------------------+------------------------------------------------
surgeon: Identity |
var(_cons) | 7339.356 4842.625 2013.824 26748.19
-----------------------------+------------------------------------------------
location: Identity |
var(_cons) | 5477.803 3445.984 1596.362 18796.7
-----------------------------+------------------------------------------------
var(Residual) | 53149.89 1783.819 49766.18 56763.67
----
Then with all pt level fixed effects (including multiple imputation)...
mi estimate: mixed caseprice c.age_dm##c.age_dm i.race_short bmi i.asa_score i.asa_emergent i.urgency c.elixsum##c.elixsum i.indication i.resident_yn || _all: R.loc || surgeon: || loc:
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
-----------------------------+------------------------------------------------
_all: Identity |
sd(R.location) | 99.45665 43.09123 42.54385 232.5042
-----------------------------+------------------------------------------------
surgeon: Identity |
sd(_cons) | 98.62659 27.50079 57.10145 170.3495
-----------------------------+------------------------------------------------
location: Identity |
sd(_cons) | 70.78175 22.57506 37.88261 132.2521
-----------------------------+------------------------------------------------
sd(Residual) | 223.3151 3.74665 216.0912 230.7805
With these results, I believe, my ICC surgeon is ~ 13%, ICC location ~ 13%, ICC for combination of surgeon and location is 33%, but not sure what I can (if anything) say about the patient level variance explained?
