Final Exam Flashcards
Estimable function
expected differences between categories of a fixed effect are estimable; the expected averages for categories of a fixed effect are not estimable
a. Expected difference between HRFI and LRFI feed intake
confounding
occurs when one explanatory variable is related to the response of interest and to another (or more than one) explanatory variable so that it is impossible to separate the effects of two explanatory variables on the response
a. Non-estimable SAS errors
b. Missing combinations of fixed effects
c. Choose variable which you think is responsible for most of the variation
a. Day and week were confounded in my study
Collinearity
occurs when two variables are highly correlated
a. As one variable increases the other variable increases proportionally
b. This won’t tell us much because the variables are so similar and correlated
c. Wastes time, $, labor
a. Hot carcass weight and loin eye area
null hypothesis
no difference between treatment groups means
a. No difference between behaviors or HRFI and LRFI pigs
alternative hypothesis
the treatment group means are different (data do not conform to null hypothesis)
b. LRFI pigs will be less stressed during HAT and NOT than HRFI pigs
precision & accuracy
a. Precision- how well repeated observations agree with one another
b. Accuracy- how well the observed value of a quantity agrees with the true value
c. One can be precise without being accurate, so accuracy may be more important
power
how many experimental units (n) are needed to detect a statistical difference
a. Best to determine power for most variable trait knowing that the power will be higher for all other traits with less variation
b. Properly designed experiments ensure power will be high enough to detect departures from the null hypothesis
c. Power influences statistical test being performed, sample size, size of experimental effects, level of error
d. Anything that increases accuracy and consistency increases power
inference
drawing conclusions about a population using sample information
a. Can only make inference relative to the population from which experimental units were chosen
experimental unit
smallest unit of sample population treated alike
a. The experimental unit is what the treatments were randomly assigned to
b. The observational unit is what the response measurements were taken on
a. Pig, pen
fixed effects
contribute variation, repeatable in other study
a. Categorical, discrete variables
b. Treatments
a. Line, diet, line*diet
random effects
selection by chance, randomly chosen from a population, not repeatable in another study
a. Animal ID
a. Week, handler
nesting
factor levels within levels of another factor
a. Season within a year within a herd
a. Handler(pig)
covariates
contributes variability that should be accounted for
a. Continuous variables (can be any real number within a reasonable range)
b. Weight, temperature
a. BW, baseline cortisol concentration
standard deviation
measure of variability for an individual observation
a. Used to indicate how widely individuals in a group vary
standard error
measure of variability of the sample mean
a. SD/√n
b. Standard deviation of the sampling distribution
c. More variation, need more observations (higher n) in order to detect differences
P value
probability that the observed relationship (between variables) or a difference (between means) in a sample occurred by chance alone
a. Measure of statistical significance
b. The degree to which the result is true, the likelihood of observing certain data given the null hypothesis is true
c. P < 0.05 indicates 5% risk of the differences occurring without treatment effect
a. P < 0.05 can reject the null hypothesis
TYpe I and Type II errors
a. Type I- accept null hypothesis when it is not true (false positive)
b. Type II- reject null hypothesis when it is actually true (false negative)
F value
ANOVA, tests variance against a null hypothesis
a. Type I sum of squares- sequential sum of squares, tests main effects for factors followed by the interaction effect
b. Type II sum of squares- tests for the presence of an effect after all other effects have been accounted for, only valid if significant interactions are included in the model (better for unbalanced data)
least square means
adjusted treatment means accounting for all other effects in the model
a. Least square means that the overall solution minimizes the sum of the squares of the errors made in the results of every single equation
b. Effects of predicted values of dependent variable for each level of the effect when all other effects are set to their mean values
