The NorFor nutrition model (http://www.norfor.info/) has been used extensively for formulating dairy rations in Scandinavia. The Nutritional Dynamics System (NDS) version of the Cornell model (http://www.rumen.it/en/ndspro) has been widely adopted by consulting nutritionists in North America. We initiated a comparison of how well NorFor and NDS predicted observed yields of energy-corrected milk (ECM) and milk true protein (MTP). Data from 5 published trials (21 diets) were entered into NorFor (version 2017.12.0.187) and NDS (version 3.9.5.01). Analyzed composition data for major feed ingredients were used to modify tabulated feed compositions in both models. NorFor reports ECM directly but total milk protein rather than MTP; MTP was computed by multiplying total milk protein by 0.95. The lower of the energy- or MP-allowable ECM and energy- or MP-allowable MTP reported by NDS were used in fitting. Data were from short-term Latin square trials; thus, yields were not corrected for BW change in either model. Predictions of ECM and MTP were regressed on observed data using SAS, including trial and diet(within trial) in the general linear model; regression parameters are in Table 1. Both models explained most of the observed variation; however, root mean square errors were smaller and coefficients of determination were greater when fitting NorFor data. Moreover, standard errors of intercepts and slopes were smaller for NorFor. Slopes from fitting ECM and MTP predictions from both models were different from 1, indicating significant slope bias. Further work using much larger databases is needed to test the relative effectiveness of the NorFor and NDS models. Table 1. Linear regression parameters from fitting NorFor and NDS predictions to observed ECM and MTP yields

Trait	Model	Estimates
R2	RMSE1	Intercept	SE1	Prob2	Slope	SE	Prob3
ECM	NorFor	0.914	0.787	6.3	3.4	0.078	0.80	0.09	<0.001
	NDS	0.855	2.168	−22.7	9.2	0.027	1.69	0.25	<0.001
MTP	NorFor	0.920	0.027	0.19	0.10	0.068	0.78	0.08	<0.001
	NDS	0.788	0.077	−0.36	0.28	0.226	1.30	0.23	<0.001

¹Root mean square error and standard error. ²Probability that intercept is different from 0. ³Probability that slope is different from 1.