242
was the event of interest (failure). Cows that had not delivered a calf before
December 22nd 2016 or 2017 were right censored (n = 28). Cows that were excluded
from the study (before the study period ended) were left censored (n = 12).
Milk samples were obtained two times weekly for progesterone assessment during the first year of the study (results reported in another publication included in this
book), such that frequent visits by the field technician enabled an accurate determination of the lactation period in 56 cows, and cessation of milk production (n = 42)
was the event of interest (failure). Cows that were still lactating at the end of the
milk sampling period were right censored (n = 14). The last day of milk sampling
was used as the day of censoring in these cows. None of the cows were left censored
before cessation of milk yield.
The calves (n = 98) were subject to body girth measurements in October,
November, December, January and March. The difference in body girth between
October and January the following year, and between October and March the following year, was used to assess associations between body girth change (weight
gain) and supplemental feeding of the dams. The age of the calf at the day of body
girth measurement in October, January and March and sex of the calves were
recorded for inclusion in statistical models.
3 Statistical Analyses
The statistical software package Stata 14 (StataCorp LLC, Texas, USA) was used in
all statistical analyses. Statistical significance was considered with a P-value less
than 0.05 in all models. Univariable Kaplan-Meier (K-M) estimators were used to
assess relationships between calving interval and supplemental feeding (feeding,
1/0). Differences in survival functions between feeding groups were tested separately for each cohort and for the two cohorts joined. Similarly, the association
between lactation period and feeding, 1/0, was assessed for the first year of the
study. Difference between the groups was assessed by the log-rank test for equality
of survivor functions.
Separate generalised multivariable linear regression models were run for the
assessment of associations between body weight attainment and the explanatory
variables: feeding, 1/0, gender (sex) and age of the calves in January and March.
Initially, univariable associations between the two outcome variables, body girth
attainment from October through January and October through March, and the
explanatory variables were explored by analysis of variance (feeding, 1/0, and sex)
or linear regression (age). Univariable associations with a P-value of less than 0.15
for the two-directional association with the outcome qualified for inclusion in a
separate multivariable linear regression analysis. A backward selection procedure
was employed, and explanatory variables with a P-value of less than 0.10 were kept
in the final model. This resulted in two models explaining the body weight attainment for the two periods. The first model assessed body weight attainment for the
period from October through January and included the explanatory variables: feedM. A. Bhatti et al.
was the event of interest (failure). Cows that had not delivered a calf before
December 22nd 2016 or 2017 were right censored (n = 28). Cows that were excluded
from the study (before the study period ended) were left censored (n = 12).
Milk samples were obtained two times weekly for progesterone assessment during the first year of the study (results reported in another publication included in this
book), such that frequent visits by the field technician enabled an accurate determination of the lactation period in 56 cows, and cessation of milk production (n = 42)
was the event of interest (failure). Cows that were still lactating at the end of the
milk sampling period were right censored (n = 14). The last day of milk sampling
was used as the day of censoring in these cows. None of the cows were left censored
before cessation of milk yield.
The calves (n = 98) were subject to body girth measurements in October,
November, December, January and March. The difference in body girth between
October and January the following year, and between October and March the following year, was used to assess associations between body girth change (weight
gain) and supplemental feeding of the dams. The age of the calf at the day of body
girth measurement in October, January and March and sex of the calves were
recorded for inclusion in statistical models.
3 Statistical Analyses
The statistical software package Stata 14 (StataCorp LLC, Texas, USA) was used in
all statistical analyses. Statistical significance was considered with a P-value less
than 0.05 in all models. Univariable Kaplan-Meier (K-M) estimators were used to
assess relationships between calving interval and supplemental feeding (feeding,
1/0). Differences in survival functions between feeding groups were tested separately for each cohort and for the two cohorts joined. Similarly, the association
between lactation period and feeding, 1/0, was assessed for the first year of the
study. Difference between the groups was assessed by the log-rank test for equality
of survivor functions.
Separate generalised multivariable linear regression models were run for the
assessment of associations between body weight attainment and the explanatory
variables: feeding, 1/0, gender (sex) and age of the calves in January and March.
Initially, univariable associations between the two outcome variables, body girth
attainment from October through January and October through March, and the
explanatory variables were explored by analysis of variance (feeding, 1/0, and sex)
or linear regression (age). Univariable associations with a P-value of less than 0.15
for the two-directional association with the outcome qualified for inclusion in a
separate multivariable linear regression analysis. A backward selection procedure
was employed, and explanatory variables with a P-value of less than 0.10 were kept
in the final model. This resulted in two models explaining the body weight attainment for the two periods. The first model assessed body weight attainment for the
period from October through January and included the explanatory variables: feedM. A. Bhatti et al.