4.11
Mean Absolute Deviation
The mean absolute deviation is also a measure of the variability of data in a sample. It is the absolute value of the average difference between the d…
The mean absolute deviation provides the absolute value of the average difference between the data values and the mean.
It is calculated as the sum of the absolute deviations from the mean divided by the sample size.
For example, three students have three, five, and seven cookies in their lunch boxes. The deviations in the number of cookies from the mean of five cookies are minus 2, zero, and two.
If one adds these deviations, the positive and negative values cancel each other out, giving a zero mean deviation, which is unhelpful. If the absolute values are added, one obtains a single non-zero value instead.
This value, when divided by the sample size, gives the mean absolute deviation.
Calculating the mean absolute deviation involves a non-algebraic modulus operation, while the standard deviation uses algebraic operations. Therefore it is not suitable for inferential statistics.
It is also a biased statistic, as the calculated mean absolute deviation of a sample does not adequately represent the population mean absolute deviation.
View the full transcript and gain access to JoVE Core videos
Q1: How is mean absolute deviation calculated from a dataset?
Mean absolute deviation is calculated by finding the average distance of data values from the mean. First, compute the sample mean. Then calculate the deviation for each value by subtracting the mean. Take absolute values of these deviations, sum them, and divide by the sample size. For example, with cupcake data of 10, 15, 8, 7, and 10, the mean is 10, absolute deviations sum to 10, yielding a mean absolute deviation of 2.
Q2: Why do we use absolute values when calculating mean absolute deviation?
Absolute values are used because positive and negative deviations cancel each other out when added together, resulting in zero. This zero value is unhelpful for measuring variability. By taking absolute values of all deviations before summing, we obtain a meaningful non-zero measure that accurately reflects the average distance of data points from the mean.
Q3: What is the difference between mean absolute deviation and standard deviation?
Mean absolute deviation uses a non-algebraic modulus operation to calculate variability, while calculating standard deviation uses algebraic operations. Because mean absolute deviation employs non-algebraic methods, it cannot be used in inferential statistics, which requires algebraic operations. Standard deviation is therefore preferred for statistical inference.
Q4: Is mean absolute deviation suitable for inferential statistics?
No, mean absolute deviation is not suitable for inferential statistics because it involves a non-algebraic modulus operation, while inferential statistics requires algebraic operations. Additionally, mean absolute deviation is a biased statistic—the calculated mean absolute deviation of a sample does not adequately represent the population mean absolute deviation.
Q5: What does it mean that mean absolute deviation is a biased statistic?
A biased statistic means the sample estimate does not accurately represent the population parameter. For mean absolute deviation, the value calculated from a sample does not adequately represent the true mean absolute deviation of the entire population. This bias limits its reliability for making population-level inferences or generalizations.
Q6: How does mean absolute deviation measure data variability?
Mean absolute deviation measures variability by quantifying the average distance of data points from the mean. It reflects how spread out values are in a dataset. A larger mean absolute deviation indicates greater variability, while a smaller value indicates data points cluster closer to the mean.
Q7: What are the main limitations of mean absolute deviation in statistical analysis?
Mean absolute deviation has two primary limitations: it cannot be used in inferential statistics due to its non-algebraic nature, and it is a biased statistic that does not reliably represent population parameters. These constraints make it less suitable than other measures for advanced statistical applications.