Please enable JavaScript to view this site.

Navigation: Data tables > Distinguishing the eight kinds of data tables

Column tables (one grouping variable)

Scroll Prev Top Next More

In a column data table, each column defines a group. These groups are defined by one scheme, perhaps "control vs. treated." You can have more than two groups in a one-way table, for example "placebo vs. low-dose vs. high-dose.

Analyses performed with Column data

Unpaired t test

Paired t test

Ratio-paired t test

Mann-Whitney test

Wilcoxon test

Kolmogorov-Smirnov test

Ordinary one-way ANOVA

Brown-Forsythe and Welch ANOVA

RM one-way ANOVA

Kruskal-Wallis test

Friedman test

One sample t and Wilcoxon test

Descriptive Statistics

Normality and Lognormality Tests

Frequency Distribution

ROC Curve

Bland-Altman

Identify Outliers

Analyze a stack of P values

 

Example of a column table

data col2

Each column defines a treatment group.

Prism does not use a separate column to enter the grouping variable. Instead the groups are defined by columns.

If data are not matched or paired, rows have no special meaning. Enter data in any order you like.

If data are from a before-after experiment with clear pairing or matching, place each set of matched results on one row. Optionally, use the row title to label it.

Blue italics values are “excluded” (ignored by analyses and graphs). Use Exclude Values.

 

Error bars

For one-way tables, Prism will calculate error values and create error bars automatically. Enter all the values for a data set in a single column.

data col3

 

 

 

 

 

© 1995-2019 GraphPad Software, LLC. All rights reserved.