Unsupervised versus Supervised Identification of Prognostic Factors in Patients with Localized Retroperitoneal Sarcoma: A Data Clustering and Mahalanobis Distance Approach

Table 3

Univariate DFA over filtered and unfiltered cases. The association with the survival status was independently tested for each of the remaining 9 variables. In the “unfiltered” condition (on the left), the valid cases depended upon the presence of missing cases (to be excluded) on that variable. In the “filtered” condition (on the right), all cases having any type of missing values (9 in total) were excluded.

	Unfiltered					Filtered
	Valid cases	# wrong	% correct	-squared	−2 log prob.	Valid cases	# wrong	% correct	-squared	−2 log prob.
Age	70	32	54.3	0.00037	96.95	61	27	55.7	−0.003	84.4
Histology	70	27	61.4	0.078	89.40	61	25	59.0	0.060	78.7
Grading	69	25	63.8	0.131	83.01	61	23	62.3	0.114	74.6
Size	68	35	48.5	−0.0023	94.25	61	31	49.2	−0.002	84.28
Response to CTRT	67	26	61.2	0.016	91.38	61	35	42.6	−0.004	84.52
Resection	70	32	54.3	0.02	95.05	61	32	47.5	0.007	83.6
DFI	70	11	84.3	0.49	49.62	61	11	82.0	0.458	45.61
Relapse pattern	66	12	81.8	0.24	69.39	61	11	82.0	0.219	65.71
First type of retreatment	68	15	77.9	0.12	83.36	61	14	77.0	0.120	74.045