Table 4. Generalized ordered logistics regression with clustered errors.
Dependent variable: survival rate (reversed).
| Independent Variables (2) | Elasticities (1) | Delta Method Standard Errors |
|---|---|---|
| Avg no. publications | 0.022 | 0.095 |
| Avg no. publications*cardiology | 0.076 | 0.066 |
| Avg no. publications*Oncology | 0.189*** | 0.032 |
| Avg. Citations | 0.267*** | 0.084 |
| Avg. Citations *cardiology | -0.107 | 0.075 |
| Avg. Citations * Oncology | 0.023*** | 0.021 |
| For-Profit | -0.284* | 0.167 |
| Oncology | 0.022 | 0.042 |
| Cardiology | 0.177** | 0.060 |
| Clinical services | 2.123* | 1.182 |
| Length of stay | 0.313 | 1.182 |
| Median age | -2.464** | 1.010 |
| Median income( | -0.712** | 0.346 |
| Physicians | -0.096 | 0.211 |
| Net profit/loss per physician | -1.883 | 1.295 |
Analyses were conducted in STATA version 13.1, using the gologit2 module [49]; an advanced version of the generalized ordered logit model [50] for ordinal dependent variables. The gologit model relaxes the proportional odds assumption and allows the effects of the explanatory variables to vary with the point at which the categories of the dependent variable are dichotomized.
N = 145, Log pseudo likelihood = -107.89, Pseudo R2 = 0.188
Wald test indicates that the final model does not violate the proportional odds/ parallel lines assumption.
***P<0.01
**P<0.05
*P<0.1
(1) Calculated at means.
(2) The variable staffed beds was omitted from this analysis because, given the sample size, the Ordered logistic regression is less resilient than the OLS regression for number of regressors.
(3) The gologit2 command, an advanced version of the generalized ordered logit model [50] for ordinal dep. Vars. The gologit model relaxes the proportional odds assumption and allows the effects of the explanatory variables to vary with the point at which the categories of the dependent variable are dichotomized. It also offers several additional powerful options such as a straightforward calculation of elasticities.