comes up with an effect size estimate and a confidence interval (low value/high value) around the estimate when an outcome is relatively unknown.

measures the relationship, or association, between two variables.

These methods account for additional information beyond historical data that might influence the variable that's being forecast.

is a good metric to remember. It is used in many statistical analyses and will be something that your analysts should know. As you can imagine, there is a good deal of research about where the p-value comes from and how it is calculated.

This measure provides a relatively real-time update to the direction of the forecasting error. Generally, this is calculated by dividing the sum of the errors by the MAD. It provides a directional percentage figure.

This category uses simple methods to predict future values,

means that the value can be quite spread out

means that the individual observations in the data are relatively close to the mean

If there are regular day of the week or monthly patterns in the busness, certain forecasting methods will be more appropriate.

As technology has advanced, providing sufficient processing power to solve larger and more complex math problems, additional complex forecasting methods have been developed.

If there is a lot of noise in the data, leaning observations jump around quite a bit and there are very few detectable patterns, it will be more difficult to use for prediction.

is used when you want to understand the relationship between the independent variables and the dependent variable or when you want to use the independent variables to predict the dependent variable.

These methods use historical data, but add more complexity for pattern detection and pattern changes, helping to account for elements like trend and seasonality.

it is a mathematical model that estimstes the relationships among two or more variables

This measure adjusts for the problem of scale just described by expressing the mean absolute error as a percentage of the total forecasted value.

When data gets sparse, as in very few historical observations are available, accuracy is impacted.

are generally used for continuous data (revenue, guest counts), whereas percentages are generally used to describe categorical data (gender, hotel class).

represents the amount of variability in the data, or how "spread out" the observations are from each other and from the mean (average).

Certain forecasting methods deal better with unusual observations or factors that influence the patterns that are not easily visible in the historical data (like oil prices, unemployment rates, or weather).

are used to draw meaningful conclusions about the population or scenario that your data represents.

are subjective, based on the opinions and judgments of experts. They are most appropriate when past data are not available (like new product forecasting), and are usually applied to intermediate or long-range decisions.

is the use of historical data to predict the direction of future trends, in business applications, forecasting is generally used to assist in the planning process, so it is most commonly used to make a prediction of revenue or demand for a product or service.

This measure is the average of the absolute value of the difference between each forecasted value and the actual value for that period.

are used to forecast future data as a function of past data. They are most appropriate to use when past numerical data is available, and when it is reasonable to assume that some of the patterns in the data will continue into the future. These methods are generally applied to short- or intermediate-range decisions

is generally based on supply and demand relationships, predicting the price sensitivity of demand, the switching behavior in the face of available alternatives, or the impact of certain financial policies on gross domestic product.

(yes/no questions) involves determining if there is enough statistical evidence to say whether something is or is not the case.

says that you are 95% confident that the answer is yes, and smaller p-values increase the confidence.

is basically a summary of the characteristics of the population represented by your data. This includes things like counts, maximum, minimum, and most frequent values, averages (mean), median, distribution (percentage breakdown), and standard deviation.