Term

Description

Example

ANOVA

A test for differences among three or more treatment methods.

ANOVA is an acronym for analysis of variance.

Chi-square

A quantitative comparison between what a researcher expected based on characteristics of subjects and what was actually observed.

The larger the Chi square value, the more probable the null hypothesis is false.

Descriptive Statistics

Common terms used to describe data.

Average, mean, median, mode, central tendency, etc.

Incidence Rate

A ratio of the number of new cases of a disease occurring in a specific time period compared with the total population during the same time period.

There were 10 new tuberculosis cases diagnosed in a town in 1976 with a population of 100 people during that time, or an incidence rate of 10%.

Non-parametric
Statistics

A statistical method that is not based on assumed parameters of a population from which subjects are selected.

Selection of subjects who live in a town where there is a high fluoride concentration in the water supply.

Nominal Scale

Categories of data that do not have a quantitative relationship to each other.

Group A: patients who floss and Group B: patients who don’t floss.

Null Hypothesis

A statement theorizing that there is no difference between (the null) groups of subjects, or factors to be studied.  The theoretical basis for a study.

"There is no difference in tooth staining between users of regular dentifrice and users of a whitening dentifrice in a 3-month study."

Ordinal Scale

Categories of data that can be rank ordered, but the difference between the categories is not specified.

Patients with stain on Tooth #8

• Group A: 20

• Group B: 15

• Group C: 10

• Group D: 9

Parametric Statistics

A statistical method that is based on assumed parameters of a population, or subpopulation from which subjects are selected.

Selection of subjects who were born and raised in a town where there is a high fluoride concentration in the water supply.

Prevalence Rate

A ratio of the number of new cases of a disease occurring at a specific time compared with the total population during the same time period.

There were 10 new tuberculosis cases diagnosed in a local clinic out of 100 people who lived in the same town at that time, or a prevalence rate of 10%.

Rate

A proportion of a group which demonstrates a specific attribute compared to the whole group.

Of 100 joggers who ran a marathon, 10 experience at least one knee injury. This is a rate of 10%.

Sign Test

A non-parametric comparison test using two paired samples and is used with data sets with an ordinal scale, or higher.

A comparison of matched pairs from each test group.  One subject from Group A with a matching subject from Group B.

Significance Level

The percentage of chance that there would be a difference in findings between groups studied even if data showed no difference in the groups.

In a survey, 50% of the sample townspeople voted for fluoridation and 50% voted against.  A significance level of 1% means that there is a 1% chance that there is a difference in these groups simply due to drawing an unrepresentative sample of subjects.

Significance Level is expressed as a "p" value.  In this example p=<0.01 means 99% of the time there is no difference between the responding groups, or that less than 1% chance that the study sample was misrepresentative.

Standard Deviation

A square root of variance used to measure the "spread" of data items around a mean value in a data set.

S = the square root of variance.

T- Test

A typical statistical test for comparing two treatment methods.  A "paired" T-test uses one sample.  An "unpaired" T-test uses two samples.

A group of 20 periodontal patients were treated with half of their jaw using a surgical method, the other half with chemical therapy.  The results are compared using a paired T-test.

Variance

A measure of the "spread" of data items around a mean value in a data set.

The larger the variance reported; the larger the spread.  Variance is reported in squared units unlike mean values.

Wilcoxson
Sign-ranked Test

A non-parametric comparison test using two matched pair samples and is used with data sets with an ordinal scale or higher.

Measures both direction of the differences like a Sign Test, but also the extent of those differences.