Example
Let's try another example. Say you're researching the effects of physical activity on health in Americans of various ages. Compute the analysis without accounting for the subjects' different ages, and it's likely there will be a really strong relationship between physical activity and overall health. (Great! Lots of jobs for professional trainers!).
However, analyze the data using ANCOVA to control for age and it's likely the relationship will get a lot weaker. That's because age is more likely to be the real cause of any relationship between health and physical activity. People are less physically active as they age because aging reduces their physical strength and flexibility. Researchers say that age "confounds" the beneficial effects of activity on health. However, this predictor wouldn't be apparent unless the test results are controlled for the age factor by analyzing the covariance of people of the same age.
Explaining analysis of covariance in language can be difficult because of its high level of complexity. In broad terms, it's a computation used when two measurement variables and two nominal or category variable are involved. As the name implies, measurement variables are things that can be measured, and are always numbers. Statistics always assumes that measurement variables could have an infinite number of potential values, but in reality, the possible values are limited by the measurement.
Meanwhile, the two nominal ("name") variables, also called "attribute variables" or "categorical variables," organize data into categories. Whereas a measurement value is always a number, a nominal value is usually a word. Some examples from biomedical statistics could include genotype (AA, Aa, or aa), sex (male or female), or condition (normal, sprained, torn or broken). Often nominal variables are used to group individuals into classes, so that other variables may be compared among the classes.
Now here's the kicker on nominal variables: one of the nominal variables is called a "hidden" nominal variable, because it's used to group the data without being named itself. For example, the "hidden" nominal variable could group measurements into pairs, while the "visible" nominal variable divides the pairs into two or more sets of data. Back to the main page
However, analyze the data using ANCOVA to control for age and it's likely the relationship will get a lot weaker. That's because age is more likely to be the real cause of any relationship between health and physical activity. People are less physically active as they age because aging reduces their physical strength and flexibility. Researchers say that age "confounds" the beneficial effects of activity on health. However, this predictor wouldn't be apparent unless the test results are controlled for the age factor by analyzing the covariance of people of the same age.
Explaining analysis of covariance in language can be difficult because of its high level of complexity. In broad terms, it's a computation used when two measurement variables and two nominal or category variable are involved. As the name implies, measurement variables are things that can be measured, and are always numbers. Statistics always assumes that measurement variables could have an infinite number of potential values, but in reality, the possible values are limited by the measurement.
Meanwhile, the two nominal ("name") variables, also called "attribute variables" or "categorical variables," organize data into categories. Whereas a measurement value is always a number, a nominal value is usually a word. Some examples from biomedical statistics could include genotype (AA, Aa, or aa), sex (male or female), or condition (normal, sprained, torn or broken). Often nominal variables are used to group individuals into classes, so that other variables may be compared among the classes.
Now here's the kicker on nominal variables: one of the nominal variables is called a "hidden" nominal variable, because it's used to group the data without being named itself. For example, the "hidden" nominal variable could group measurements into pairs, while the "visible" nominal variable divides the pairs into two or more sets of data. Back to the main page