Statistics 110E -- Statistical and Data Analysis-Psychology/Biological Sciences
Statistic 110 Lab 6
JMP IN Topics
- 2 x 2 Contingency Tables
To illustrate the commands for analyzing a contingency table, we will
use the data from HW problem 17, Chapter 12.
Preliminaries for Exercise 12:17
- Start up JMP IN
- Double-Click on "0 Rows" in the upper left corner of the spreadsheet. In
the Pop-Up window, specify the number of rows to add, 4, corresponding
to the 4 cells in the table. Click on Add
or simply hit Enter.
- Add columns for the two categorical variables: "Helped" with
possible values "Yes" or "No" and
"Observer" with possible values "Male" or "Female". To convert a column
to a categorical variable, select the column, and then under the Cols
menu, select Column Info. In the pop-up window, change the data type to
Character. The left box at the top of the column should have a "N" for
Nominal data. The four rows should correspond to all possible
combinations of the two variables making up the 4 cells in the table.
- Add a numerical column for the cell counts, "Counts". Enter the
corresponding counts for each row. Click on the right-hand box at the
top of the column, and select "F" for Freq or frequency.
This form of data entry assumes that the data have already been
summarized into a 2 x 2 table. If you have raw data, i.e. the categories
for each individual observation, you would have n rows with just the 2
columns for the categories, and no "Count" variable (i.e. counts would
all be one).
Analyses
- Go to the Analyze menu and select "Fit Y by X"
- Select "Helped" as the "Y" and select "Observer" as the "X".
"Counts" should appear in the Freq field.
- The Analysis type should be "Contingency Table, Crosstab"
- Click on OK.
Output
- The plot at the top is a Mosaic plot. It shows the proportions
that helped picked up/did not help pick up pencils for Males and
Females. (basically a stacked Bar-graph). This provides a very nice
visual impression of the differences between the two groups.
- The second box contains the Cross-tabulated table. Verify that the
numbers are correctly entered! To get the Expected Counts and Chi-Square
values for each cell and percentages, click on the right arrow next to
"Crosstabs" and select the corresponding items to toggle them on (there
should be a check next to them).
- The last box has the output for testing if there is a statistical
relationship. The only row that we need for now is the row
corresponding to "Pearson". The column labeled "ChiSquare" has the
chi-squared summary measure. Verify that it is the sum of the Chi^2
values given in 2 or what you get by hand. The Prob>ChiSq column has the
corresponding p-value, which is the proportion of other samples that
would have a ChiSquare summary statistic as large or larger than what we
observed based on chance alone (i.e. when there really is no relationship
between gender and helping).
- To test if the relationship is statistically significant, compare
the ChiSquare value to 3.84. If the ChiSquare summary statistic is
greater than 3.84, then there is a statistically significant
association between the gender of the observer and whether they would
help pick up the pencils. You can also use the p-value, so if the
p-value is less than 0.05, then there is a statistically significant
association.
- Save your output to a journal file and clean up any extraneous
information (i.e. delete output that we have not discussed such as
Fisher's exact test).