How to Lie with Statistics

A bar chart is a graph used to represent comparisons in the numbers of a given sample. Each bar represents a discreet category. The height of each bar—or length in the case of a horizontal bar chart—represents that category’s given value.

An index number is a number in statistics representing the change between two points. It is calculated by comparing the quantity of a thing with a base value. Index numbers such as price, which change over time, often appear in economics studies. Darrell Huff notes that they are often manipulated to suit political and commercial agendas.

A line graph is a type of graph that plots a change over time. Individual data points are placed on the graph according to their x- and y-axis coordinates. Once all the points are set, a line connects them to show the trajectory of the change.

The mean is one of the three main types of statistical averages, along with the median and the mode. It is the numerical average of a data set. To find the mean, a statistician adds all the numbers in a sample and then divides by the total number in that sample. Huff notes that the type of average used is often manipulated to distort data.

The median is the second of the three main types of statistical average, along with the mean and the mode. The median is the central number in a dataset. It can be found by arranging all the examples in a sample from smallest to largest and then locating the middle number. The average is taken from the two center-most points in samples with an even number of points.

The mode is a third type of statistical average, along with the mean and the median. The mode is the number that occurs most frequently within a given sample. To find the mode, the statistician counts the occurrence of each number within the sample. The number with the highest occurrence is the mode. The sample has no mode in cases where no number appears more than the others.

The ordinate is a term used for coordinates on a line graph, in opposition to the abscissa. The book uses the simplified meaning of the term, wherein ordinate refers to a given point’s coordinates on a graph’s y-axis.

The percentile is a score that indicates the percentage of examples in a distribution that falls below a certain range. It exists on a scale of 100. Percentile is not the same as the percentage. A percentage is an individual score, while a percentile presents a relative rank compared to others in the set. Percentiles are often used in standardized testing to rank students. Huff notes that this can often lead to deceptive interpretations, as students’ results tend to form a bell curve; thus, the difference between students in the 40th to 60th percentile may be minimal, but the difference in performance between upper-tier percentiles is much more pronounced.

A pictorial graph is a graph that uses pictures or other visual representations of the studied population to represent statistical data. It is often used for comparisons, and the size of each picture changes to represent the difference. They are more common in examples meant for general audiences than in academic publishing.

In statistics, the population refers to the collection of entities a statistician chooses to measure. It can refer to a population in the everyday use of the word, such as a population of people. However, it can also refer to any group of individuals or objects used in statistical analysis.

Post hoc is the shortened version of the longer Latin phrase post hoc, ergo propter hoc, meaning “after this, therefore because of it.” It refers to the fallacy of thinking that a sequential relationship between two things also implies a causal relationship. Huff uses it in the text to indicate any instance of a false belief that correlation equals causation, even when the sequential relationship cannot be determined.

In statistics, sampling refers to a statistician’s method to measure a population. A sample is the portion of the population measured in a given statistic. Accurate results require random samples that don’t favor certain traits or subsections over others. If samples are not random, introduced biases can skew the resulting data. Due to the difficulty of acquiring actual random samples, statisticians take a stratified random sample instead. To stratify means to divide the total population into subgroups based on traits. A stratified random sample selects from these subgroups.