Statistical terms

Many statistical terms appear in IBM® Cognos Analytics with Watson. It is important to understand the definitions of these terms as they apply to Cognos Analytics.

Adjusted count R-squared
Adjusted count R-squared is a measure of improvement from the constant model to the fitted model.
Adjusted count R-squared category filter
In some calculations of the adjusted R² accuracy, data for categories with low total counts are not used in the calculation.
Adjusted R squared
Adjusted R² is a corrected goodness-of-fit (model accuracy) measure for linear models. It identifies the percentage of variance in the target field that is explained by the input or inputs.
Categorical
Categorical is a type of measurement level. A categorical field is a field with a limited number of distinct values. The values represent a type or kind.
Chi-square value
The chi-square value is a value on the chi-square distribution. A statistical test generates a chi-square value. The value specifies the probability of obtaining the chi-square value from chance. If the probability is less than the significance level, the test is judged to be statistically significant.
Chi-square adjustment for sparse data
In some applications of the chi-square statistical test, the calculation of the chi-square value is adjusted by a small factor.
Continuous
Continuous is a type of measurement level. A continuous field is measured numerically so that distance comparisons between values are appropriate.
Cramér's V
Cramér’s V is an effect size measurement for the chi-square test of independence. It measures how strongly two categorical fields are associated.
Decision rule
A decision rule is a set of conditions that classify records. The rule predicts an outcome in the target field.
Departure from overall distribution
Departure from overall distribution is an effect size measurement for the distribution test. It measures the extent to which a conditional distribution does not match the marginal distribution of the target field.
F value
The F value is a value on the F distribution. Various statistical tests generate an F value. The value can be used to determine whether the test is statistically significant.
Frequency
Frequency is another word for count. It measures the total number of records or the number of records in a specific group. In other words, frequency measures how often a value occurs.
Frequency weights
A numeric field can be used to specify frequency weights. Each record in the data may represent more than a single observations. Corresponding observation counts are frequency weights needed for building the model.
Gini impurity measure
Gini impurity is a measure used for generating classification trees. It provides more information on the distribution of data per node than the classification accuracy used for reporting the tree accuracy.
Key driver
A key driver is an input field or combination of input fields that has a statistically significant effect on a target field. This effect is more than a mere association because the effect is directed. The input field or fields are influencing the values in the target.
Modified z score
The modified z score is a standardized score that measures outlier strength or how much a particular score differs from the typical score. Using standard deviation units, it approximates the difference of the score from the median.
Normal distribution
The normal distribution is a theoretical distribution of values. It is often called the bell curve because the visual representation of this distribution resembles the shape of a bell. It is theoretical because its frequency distribution is derived from a formula rather than the observation of actual data.
Overall distribution
The overall distribution is the distribution of records in the target field before any grouping with the input field or fields.
Regression weights
An optional numeric field can be used to specify regression weights for a multiple regression model. Regression weight for a given record corresponds to influence of the observation on the computed model parameters. An example of regression weight field is size field in a bubble visualization.
R2
R² measures how well a regression model fits the actual data. In other words, it is a measure of the overall accuracy of the model. R squared is also known as the coefficient of determination.
Significance level
The significance level is the cutoff for judging a result as statistically significant. If the significance value is less than the significance level, the result is judged to be statistically significant. The significance level is also known as the alpha level.
Significance value
The significance value, or p value, is the probability that a result occurred by chance. The significance value is compared to a predetermined cutoff (the significance level) to determine whether a test is statistically significant. If the significance value is less than the significance level (by default, 0.05), the test is judged to be statistically significant.
Standard error
The standard error is an estimate of how much the value of a test statistic varies from sample to sample. It is a measure of the uncertainty of the test statistic. Standard error might be abbreviated as std. error.
Statistical distribution
A statistical distribution, or probability distribution, describes how values are distributed for a field. In other words, the statistical distribution shows which values are common and uncommon.
Statistical significance
Statistical significance is a judgment of whether a result occurs because of chance. When a result is statistically significant, we mean that it is unlikely that the result occurs because of chance or random fluctuation.
Studentized residuals
Studentized residuals allow comparison of differences between observed and predicted target values in a regression model across different predictor values. They can also be compared against known distributions to assess the residual size.
t value
A t value is the result of a statistical test. The value is located on the Student’s t distribution that is appropriate for the degrees of freedom. The location specifies the probability of obtaining the t value from chance. If the probability is less than the significance level, the result is judged to be statistically significant

Parent topic: Dashboards