Livestock Research for Rural Development 14 (5) 2002

Citation of this paper

Correlations of part and total lactations, and the prediction of lactation milk yield in Venezuelan dual purpose cows hand milked with calf at foot

José Alfredo Llamozas* and Lucia Vaccaro

Universidad Central de Venezuela, Facultad de Agronomía, Maracay, Venezuela
* Present address: c/o Banco Provincial, Caracas, Venezuela;


The object of this study was to explore the use of  part lactations for predicting total lactation milk yield (Pt) in dual purpose cows milked by hand with calf at foot. Pt, corrected to 244 days, was obtained from daily records of 902 lactations of Bos taurus x Bos indicus crossbred cows on three extensive, lowland  farms in central Venezuela. Part yields (Pp ) were calculated over days 1-30, 31-60, 61-90, 91-120, 121-150 and 151-180, and part accumulated yields  (Pac) over days 1-60, 1-90, 1-120, 1-150 and  1-180. Data were classified by farm (F; 1,2,3), year/season of calving (YSC; 1990-1992; wet, dry), calving number (CN; 1, 2+3, > 3),  breed group (BG; < 50, 50,  > 50% Bos taurus inheritance) and the interval between calving and the start of milking (ICM; 0-10, 11-20, 21-30, >30 days).

The overall correlations of Pt  with Pp  rose significantly (P < 0.01) from 0.57** (1-30 days) to 0.85**(121-150 days), and  with Pac  from 0.57** (1-30 days) to 0.94** (1-180 days). Significant differences  were found between the correlations within periods, principally due to YSC (Pp  and Pac ) and ICM (Pac ), but not BG. The effects of  F  and CN were not consistent. The highest  correlations were obtained from  Pp  91-120 (0.83**) and 121-150 (0.85**) days, and from Pac 1-120 (0.83**), 1-150 (0.89**)  and 1-180 (0.94**) days, but the Pp  allowed reduction of the recording period to a single month and generated correlations unaffected by ICM. Using linear regression equations derived from Pp 91-120 days, Pt   was estimated with  a mean error equivalent to 15% Pt , with 38% of errors within 10%. The corresponding  values for Pp 12l-150 days were 12% and 51%.

It was concluded that daily recording over the period 121-150 days  could provide a  simple option for helping farmers identify outstandingly good and poor cows,  rationalize culling and start the urgent process of improving the efficiency of these traditional systems.

Key words: Part lactations, correlations, prediction of lactation yield, dual purpose cows, traditional milking system


About 90% of milk produced in Venezuela comes from dual purpose farms, many of which are still in the hands of traditional producers with rudimentary management. Milk recording is essential for improving the efficiency of these herds and for any selection programme. The main limitation is the difficulty of  obtaining reliable records at farm level.  Florio et al (1998) showed that monthly recording, which is generalized in improved systems, leads to very large errors in the prediction of milk yields, under traditional conditions of management where cows are hand milked with calf suckling. More frequent recording is not an attractive alternative  for these farms, and other options which are more likely to be adopted by producers need to be explored. The objective of this paper was to calculate the correlations between part lactations and total yields and determine the possibility  of using the former to predict total yield.

Materials and Methods

The records came from 902 lactations of 615 cows on three farms which cooperate with a genetic improvement research project, based at the Universidad Central de Venezuela.

The farms are located in the state of Guárico at 130-150 m a.s.l., with annual temperatures of 26-270 C, annual mean rainfall of 940-1230 mm and 6-8 months dry period each year. The production systems combine dual purpose cattle with maize and sorghum production. The herds were kept on grazing, mainly natural grass species and  cereal stubble, with little or no supplementary feed. Cows were milked by hand once daily. Calves sucked for a few seconds  to stimulate letdown, and  took the residual milk at the end of milking. However, unquantifiable variation existed in the amount left by the milker, depending on the age and physical condition of the calf. Milking started at various intervals after calving, depending mainly on the nutritional status of the cows, and was often delayed when they calved in severe drought or very unfavourable conditions.

The records refer to daily  milk yields, obtained by farm staff trained by the project technician who supervised the work during monthly visits. A financial incentive was offered to the recorder to help ensure continuity and reliability of the records. All lactations of one or more day’s duration from cows which calved in the period 1990-1992 were included. Of a total of 1003 lactation records, 101  were omitted  because data were missing over more than five consecutive days. When fewer than five days’ data were missing, yields were estimated by taking the arithmetic mean of the two values available immediately before and after the missing period. By summing the daily records available from the first day of milking, part lactation (Pp ) yields were obtained for the periods 1-30, 31-60, 61-90, 91-120, 121-150 and 151-180 days, and part accumulated (Pac ) yields for 1-60, 1-90, 1-120, 1-150 and 1-180 days. The number of observations decreased from 902 in the first period to 678 in the last ones (151-180, 1-180 days) due to the number of cows which dried off in successive months. Total lactation yield (Pt ) was adjusted to 244 days, following the practice of the official Venezuelan recording scheme, by summing  daily yields to day 244, or to the last day of milking if cows went dry beforehand.

The data were classified according to the following factors, which had been identified as significant sources of variation in milk yield in earlier analyses of data from these and neighbouring farms: farm, year and season of calving, calving number and  breed group (Vaccaro et al 1996). They were also classified by the number of days which elapsed between calving and the start of milking.  The subclasses generated and the numbers of observations available are shown in Table 1.

Table 1. Classification of records and numbers of observations, according to different sources of variation
Source of  variation












Year / season of calving





















Calving number







2 + 3



> 3


Breed group



(Fraction B.taurus)

< 50%






> 50%


Interval calving - 1st milking (days)




0 - 10



11 - 20



21 - 30



> 30


Correlation coefficients were calculated between Pt and Pp, as well as  Pac, using the data  as a whole (global correlations) and then for each  subclass shown in Table 1. The values of the coefficients were compared across periods and between subclasses, within periods, using tests of homogeneity (Steel and Torrie 1960). The Pp  and the Pac  which generated the highest global correlations were selected as promising for predictive purposes. From these,  the periods which offered the best options in terms of reduced recording time and high R²  were further selected, and Pt  was estimated using the corresponding linear regression equations. The errors of prediction were obtained from the difference between the real and estimated Pt for each lactation record, and the means and frequency distributions were calculated for each selected Pp and Pac. The means were calculated using the absolute values of the errors (ignoring the sign) so that positive values would not cancel negative ones. They were expressed both in kilograms of milk and as a percentage of the true Pt .  


Unaccumulated part yields (Pp)

The global coefficients of correlation obtained between the Pp  and Pt  are shown in Table 2. They were all  different from zero (P < 0.01). They rose significantly from 0.57 in the first period (1-30 days) to 0.83 in the fourth (91-120 days), remained stable in the fifth  (0.85) and decreased significantly to 0.75 in the sixth (151-180 days). The fourth and fifth periods (91-120 and 121-150) were the most promising for predictive purposes (R² = 0.69 – 0.72). Table 2 also shows the minimum and maximum values obtained, according to subclass, within each of the sources of variation considered. All the correlations were highly significant (P < 0.01).

Table 2. Global and extreme correlation coefficients between total and un-accumulated part lactation milk yields (Pp)



Correlation coefficients*



1 - 30


31 - 60


61 - 90


91 - 120


121 - 150


151 - 180














0.57A x


0.68B x


0.78C x


0.83D x


0.85D x


0.75C x
Extreme values according to:













0.51 - 0.64


0.64 - 0.73


0.75 - 0.80


0.76j - 0.84k


0.82 - 0.86


0.66j - 0.85k y
Calving number


0.49 - 0.62


0.66j - 0.76k


0.76 - 0.80


0.78 - 0.85


0.81 - 0.87


0.69 - 0.79
Year / season of calving 0.52j- 0.74k y


0.63j - 0.85k y


0.72j - 0.88k y


0.77j - 0.93k y


0.79j - 0.92k y


0.69j - 0.86k y
Interval calving - 1st milking 0.42 - 0.61


0.62j - 0.73k


0.69j - 0.81k


0.76 - 0.85


0.79 - 0.87


0.66 - 0.80
Breed group


0.54 - 0.59




0.78 - 0.79


0.83 - 0.84


0.83 - 0.86


0.75 - 0.76
ABCD Global correlations with different letters are significantly different  (P<0.01)
jk Within source of variation and period, extreme values with different letters are significantly different (P<0.01)
xy  Extreme values and respective global values with different letters are significantly different (P<0.01)
* P<0.01

The minimum values obtained according to year/season of calving were significantly lower (P < 0.01) than the maximum values in every recording period. In five of the six year/seasons considered, the low values corresponded to lactations initiated in the dry season. None of the minimum values differed significantly from the corresponding global correlation. However, all the maximum values exceeded the respective global correlations, reaching over 0.90 in Pp 91-120 and 121-150 days.  

Significant differences were also found between the extreme values, according to farm, calving number and calving-milking interval in one or two of the recording periods (Table 2), but none of extreme values differed significantly from the respective global value (except for Pp  151-180 days, due to farm), and there was no consistent pattern as to the period affected. Breed group was the only factor considered which did not affect the correlations in any of the periods studied.

Accumulated part yields (Pac )

The correlations between  Pac and  Pt, shown in Table 3, were all highly significant (P <0.01), and increased in every period from 0.57 to 0.94 (P < 0.01).  The last three Pac  generated correlations higher than 0.80 (R² = 0.69 – 0.88).

Table 3. Global and extreme correlation coefficients between total and accumulated part lactation milk yields (Pac)



Correlation coefficients*



 1 - 30


 1 - 60


 1 - 90


 1 - 120


 1 - 150


 1 - 180


902 901










0.57A x 0.71B x


0.76B x


0.83C x


0.89D x


0.94E x
Extreme values according to:












0.51 - 0.64 0.66 - 0.72


0.74 - 0.79


0.80 - 0.85


0.87 - 0.90


0.91 - 0.95
Calving number


0.49 - 0.62 0.63j - 0.74k


0.74 - 0.80


0.81 - 0.86


0.89 - 0.90


0.94 - 0.95
Year / season of calving 0.52j - 0.74k y 0.61j - 0.82k y


0.72j - 0.87k y


0.79j - 0.91k y


0.87j - 0.94k y


0.91j - 0.97k y
Interval calving - 1st milking 0.42 - 0.61 0.54j - 0.73k


0.61j y - 0.79k


0.71j y - 0.85k


0.79j y  - 0.91k


0.87j y - 0.95k
Breed group


0.54 - 0.59


0.67 - 0.69


0.76 - 0.78




0.89 - 0.90


0.94 - 0.95
ABCD Global correlations with different letters are significantly different  (P<0.01)
jk Within source of variation and period, extreme values with different letters are significantly different (P<0.01)
xy  Extreme values and respective global values with different letters are significantly different (P<0.01)

* P<0.01

The extreme values obtained according to subclass (Table 3) show significant differences between them in every period due to year/season of calving and, after the period 1-60 days, due to calving-milking interval. The minimum values were associated with dry season calvings and with lactations where milking started > 30 days after calving. The highest correlations were found when milking started 1-10 or 11-20 days postpartum. Neither farm nor breed group influenced the correlations and calving number only had an effect on the values obtained for Pac 1-60 days.

Prediction of Pt  

Two Pp and three Pac  generated correlation coefficients above 0.80, with values of R² in the range 0.69 – 0.88. The two Pp   (91-120 and 121-150 days) were selected to predict Pt because they would imply recording yield for only 30 days, and the correlations were very similar to those obtained when milk was recorded continuously from the start of lactation (Pac 1-120 and 1-150 days).  Table 4 shows the true Pt  for the cows with records available in the two periods, the regression coefficients and the errors obtained by comparing predicted and true yields.

The mean errors, 130.8 ± 3.5 kg  (91-120 days) and 108.4 ± 3.4 kg (121-150 days) represent 15% and 12% of the mean Pt , respectively.  The extremes reached + 60% and –95% (91-120 days) and + 68% and – 79% (121-150 days) of the  average Pt .  However, the frequency distribution  shows that 19-28% of the errors were below 5%, 38-51% below 10% and 72-82% within 20% of the true lactation yield, Pt  (Table 4).

Table 4.  Regression coefficients and errors of prediction of total lactation yield from yields in 91-120 and 121-150 days

Period (days)









True milk yield kg/lactation (Pt)1




Regression coefficient (kg/kg)







Prediction errors
Mean error





   kg (± S.E)


131 ± 3.5


108 ± 3.4
   % Pt¹




Extreme errors (kg)













Distribution of errors (%)





   < 5% Pt




   < 10% Pt




   < 20% Pt




¹ Corrected to  244 days






The correlations reported in this study appear to be the only values in  the literature which refer to the traditional system of milking by hand with calf suckling, although milking  procedure is not always described in other studies. The values for Pp and  Pac  obtained in the period 1-90 days (0.57 – 0.78) are at the lower limit of those published in temperate climates (Van Vleck and Henderson 1961) as well as the tropics (Sharma et al 1983; Nobre et al 1985; Cardoso et al 1988; Ahunu and Danbaro 1992; Gahlot et al 1994, 1996). However, the values obtained for the second semester (0.83- 0.94) coincide closely with other results, except for the value obtained for Pp  151-180 days (0.75) which was abnormally low. This may be related to the short average lactation length in these cattle (235 ± 2.5 days, Florio et al 1998) and the fact that, as shown by the decrease in the numbers of observations for successive Pp,  25% of the cows were already dry by the sixth month.

The first three months of lactation  would be the most convenient for restricted milk recording from the practical point of view. But the correlations with total yield were less than 0.80, both for Pp and  Pac  up to the first 90 days, and only 32-61% of the variation in Pt  was explained by variation in yield during this period, limiting the usefulness of either Pp or Pac  for predictive purposes. The first month showed the lowest correlation with Pt, coinciding with evidence from other studies (Van Vleck and Henderson 1961; Sharma et al 1983; Nobre et al 1985; Cardoso et al 1988; Ahunu and Danbaro 1992; Gahlot et al 1994, 1996).

In order to reach correlations above 0.80, the values for Pac  show that one option would be  to record yield daily up to at least 120 days from the start of milking. Although this is some improvement over recording yield for the whole lactation, it would seem more practical to use the Pp  91-120 or 121-150 days, which imply recording just for a specific month, and give correlation coefficients (0.83 and 0.85) about as high as when yield is recorded continuously from the start of milking  (Pac: 0.83 and 0.89).

In addition, the correlations obtained with Pp were more  homogeneous than  Pac  and did not vary according to the interval between calving and the start of milking.  The variation caused by this factor in the Pac  correlations continued to appear up  to the sixth month, and the minimum values obtained in every period were significantly lower than the global values. The effect was due to the lactations in which milking started at >30 days after calving, which is feature of management peculiar to extensive farms and the results are therefore not extrapolable to other systems.

For the two Pp  selected as promising for predictive purposes (91-120 and 121-150 days), significant differences were found between the extreme values due to farm and year/season of calving in the first case, and only due to year/season of calving in the second. However, the minimum values according to these factors were not significantly different from the global correlations for the respective periods in either case, suggesting that the global values could apply generally. The lowest correlations due to year/season of calving occurred in the dry season of the first year of the study, probably in association with the severe shortage of feed and water due to exceptional drought that year.

Pp 121-150 days generated slightly more precise results than Pp  91-120 days in predicting Pt, with mean errors equivalent  to 12% and 15% Pt , respectively.  The mean of 12% is only somewhat higher than the 7-8% obtained by Ríos (1991) and Florio et al (1998) using monthly recording  throughout the lactation in these and neighbouring herds.   Although the magnitude of the extreme errors is of concern, the distribution shows that 28% of the errors obtained with Pp 121-150 days did not exceed 5% Pt  and are thus equivalent to the errors obtained by weighing at two-week intervals during the whole lactation (Ríos 1991; Florio et al 1998). Only 18% errors exceeded 20% Pt  and the fact recording milk for a single month could provide estimates of  Pt  with 51% errors within 10% is of notable practical interest for systems where the most frequent alternative is not to record at all.  In addition, the relative homogeneity of the correlations using Pp 121-150 days gives certain robustness to the method, because factors such as farm, breed group, month and number of calving are usually significant sources of error when Pt  is estimated from weighings made monthly or at other regular intervals during the lactation in these herds (Ferrer et al 1991; Ríos 1991; Florio et al 1998).

The extensive milk production systems which were the subject of this study are unlikely to be competitive in the long term, as the process of globalization proceeds (Ordóñez 1998). However, at present they play an important role in Venezuela, producing cheese to satisfy a strong local demand, generating daily income to cover operating costs and acting as a stabilizing buffer in the high-risk enterprise of cereal production. At the same time, the introduction of milking is a first step in the process of intensification of many beef herds in the region. These farms will be forced to intensify or to change to other agricultural enterprises in the medium term, and it is therefore urgent to provide easily adoptable  technologies, such as simple milk recording, to help them make the necessary changes. In the herds under study, culling for low milk production accounted for 9-36% of all losses, according to farm (Vaccaro and Florio 2002). This shows that a reasonable intensity of selection for milk yield can be achieved in these herds, provided they have a milk recording system, even though low rates of fertility, growth and survival limit the opportunities for culling in many cases. On the other hand, it has also been shown that the variation in milk production between individual cows within these herds is very high: the average yield of the best and worst cows in a sample of similar herds reached up to +270% and –100%, respectively, of the contemporary means (Vaccaro et al 1992).  Whatever the genetic component of this variation and long term consequences of selection, differences in production of  this magnitude would seem to justify the investment in milk recording, if only to identify  the poorest ones for culling. In the present case, 13% of the cows were dry before the fifth month of lactation. These would be candidates for an initial culling, because the correlation between lactation length and yield, as well as the repeatability of yield, is  high (Velásquez and Vaccaro 1993; Vaccaro et al 1996). In those herds which have sufficient replacements to allow further culling, milk recording during days 121-150  in the cows which survive the initial selection could be a more attractive option than sampling the whole herd at regular intervals from the start of lactation. In spite of some risk of error, the method would seem to meet satisfactorily the main objectives which are to identify outstandingly good and bad cows, rationalize the culling process, and  encourage farmers to take a more active interest in the improvement of their herds.


The authors wish to thank the farmers who generously lent their herds for use in this study. They also acknowledge gratefully the contributions of Humberto Mejías, Jazmín Florio and Jenny López. The project was financed by IDRC (Canada) and the Universidad Central de Venezuela (CDCH Proyecto 01.36.4252.99).


Ahunu B K and Danbaro G 1992 Prediction of lactation yield in crossbred cattle. Indian Journal of Dairy Science 42(5):241-245

Cardoso V L, Benintendi P, Freitas  M A and Gonçalves A M 1988 Estudo preliminar sobre a utilição de produções parciais de leite de vacas Gir de seleção leitera como indicativo do seu potencial productivo. XI Reunión Latinoamericana de Producción Animal, La Habana, G 20 (Abstract)

Ferrer J, Vaccaro L and Vaccaro R 1991 Estimación de la producción de leche en bovinos de doble propósito a partir de muestras tomadas de distintos intervalos de tiempo. Informe Anual del Instituto de Producción Animal 1990-1991. Universidad Central de Venezuela. Facultad de Agronomía. pp 107-108 (Abstract)

Florio J, Vaccaro L, Pérez A and Mejías H 1998 Errores de predicción de la producción de leche por lactancia a partir de pesajes a diferentes intervalos de tiempo en vacas de doble propósito. Livestock Research for Rural Development Volume 10 Number 1:

Gahlot C G, Gahlot R S and Pant K P  1994  Performance of Rathi cows in arid Western Rajasthan. Indian Journal of  Animal Science 64(11):1291-1295

Gahlot  C G,  Jairath L K and Gahlot  R  S  1996  Cumulative milk yield of Rathi and Rathi x Red Dane cows. Indian Journal Dairy Science 49(1):12-16

Nobre  P R,  Milagres  J C,  de Silva  A E and Ludwing A  1985  Lactation curves in the dairy herd at the University of Viscosa, Minas Gerais State. Revista da Sociedade Brasileira da Zootecnia 14(4):493-500.

Ordóñez J A 1998 Competividad del doble propósito en el trópico Latinoamericano. Archivos Latinoamericanos de Producción Animal 6(1) suplemento 1:33-44

Ríos L 1991 Estimación de la producción parcial y total de leche por lactancia en bovinos de doble propósito, en base a muestras tomadas a diferentes intervalos de tiempo. Thesis Ing. Agr. Universidad Central de Venezuela. Facultad de Agronomía. Departamento de Zootecnia. Maracay, Venezuela. 37 p

Sharma  L  D,  Thapan  P C  and  Chaudhry A  L  1983  Part-lactation as measure of breeding value in Hariana cattle. Indian Journal of Dairy Science 36(4):333-337

Steel  R  G  and Torrie J H  1960 Principles and Procedures of Statistics. Mc Graw-Hill Book Company, Inc. USA. 473 p

Vaccaro L and  Florio J 2002  Herd life, stayability to fourth calving and reasons for disposal of Bos taurus x Bos indicus cows on dual purpose farms in Venezuela. Livestock Research for Rural Development Volume 14 Number 3:

Vaccaro L, Pérez  A and Mejías H 1996 Correlations between first and successive records on four traits as a basis for culling dual purpose cattle. Livestock Research for Rural Development 8 (2): 55-59

Vaccaro L, Vaccaro R, Verde O, Alvarez  R, Mejías H,  Ríos L and  Romero E 1992  Características productivas para evaluar explotaciones y vacas en sistemas de doble propósito. Turrialba  42:14-22

Velásquez  E  and  Vaccaro L  1993 Duración de lactancia y factores que la afectan en vacas de doble propósito. Informe Anual del Instituto de Producción Animal 1992-1993. Universidad Central de Venezuela. Facultad de Agronomía. p 55-57 (Abstract)

Van Vleck L D  and  Henderson C R 1961 Estimates of genetic parameters of some functions of  part lactation milk records. Journal of Dairy Science 44(6):1073-1084

Received 8 August 2002

Go to top