Instructions for SAS JMP

Reading the Data:

To import the data into JMP, go to the "File" menu and choose the "Import" option. This should open the import window:

Find the data file (you should have already saved the data file to a local disk) and highlight it. Then make sure to click the "Labels" option since the first row of the data file has the names of the variables.

When this is done, click "Import". Then JMP should open a data window that looks like this:

 

 

Descriptive Statistics:

Notice there is a second box at the top of each column, to the right of the one used to specify data type. These are used to declare the type of variable each column will be. In the same manner in which you specified the data types, use the mouse to specify your independent (X) and dependent (Y) variables. In this design, active is our only independent variable, while score1, score2, and change are our dependent variables. When you finish, the top of the columns should look like the following picture:

Now go to the "Analyze" option on the menu bar. Then choose "Fit Y by X" from the pull down window. You should then get the "Y by X" window (shown below).

Go to the arrow button beside "Analysis" below the plot for Score1 (notice the location of the mouse pointer in the above picture). Then from the pop up menu, select "Quantiles" and "Means and Std Deviations". This will put two boxes below the plot of score1 and it will also automatically generate quantile plots, super imposed on the data. These two boxes will look like the results shown on the JMP analysis page. To simply generate a quantile plot, click here.

 

Detailed Descriptive Statistics:

You can also get the above Descriptive statistics by going to the "Analyze" option on the menu bar, and then selecting "Distribution of Y" from the pull down menu. Then "Quantiles" and "Moments" are listed below the histograms for each variables. To get the detailed descriptive statistics such as skewness and the like, click on the button beside "Moments", shown below with the mouse pointer, and select "More Moments"


See results

 

Quantile Plots:

Since Score1 is basically a pre-treatment measure, there is no real reason to perform statistical analysis on this variable. before performing any analysis, make sure you go to the box on the right hand side at the top of the "score1" column and click on the "Y" and choose "None" When you are finished, the top of the data table should appear as below:

 

Next go to the "Analyze" option on the menu bar. Then choose "Fit Y by X" from the pull down window. You should then get the "Y by X" window (shown below).

For instructional purposes, the picture shows the Quantile plots already created for "Change By active" but not for "Score2 By active" to give a before and after picture. Go to the "Display" option for "Score2 By active" (Shown by the pointer in the picture) . Then un-select "Show points", and select "Quantile Boxes" and "Means Dots, Error Bars". This has already been done in the picture above for "Change By active".

See results

Normal Probability Plots 

Again, be sure the right hand box at the top of the column for score1 is blank by clicking it and selecting "None" from the pull down menu. Then go to the "Analyze" option in the menu bar. Then select "Distribution of Y" from the pull down menu. If you have already assigned variables as X and Y as described above, you should get the "Distribution" window. It will show a histogram, Quantiles and moments for Score2 and change (shown below).

Notice in the lower left hand corner of the "Distribution" window, there is a little box with a check in it. (The mouse pointer is on it in the picture.) Click on this check mark and choose "Normal Quantile Plot" from the pop up menu. You can also remove the histogram, quantiles and other sections in the window by going to the various items in the pop up menu that have a check mark beside them and clicking on them to remove the check mark. If you do so, you get the results reported on the JMP analysis web page.

 

T Test

Once again, be sure the right hand box at the top of the column for score1 is blank by clicking it and selecting "None" from the pull down menu. Then go to the "Analyze" option on the menu bar. Then choose "Fit Y by X" from the pull down window. You should then get the "Y by X" window (shown below).

Go to the button beside "Analysis" located below each plot (shown in the picture above by the location of the mouse pointer. Then, from the pull down menu, select "Means, Anova/t-test". This does both a one-way ANOVA (2x2) and a t-test. Since comparing means is usually a simple t-test, we only reported the t-test for this case study.

See results (t-test only)

 

ANOVA

First, we have to assign the data correctly. Make sure the boxes at the top right of the columns for score1 and score2 has "Y" and the box at the top right of the column for change is blank (recall that you do this by selecting "None" from the pull down menu after clicking on the box.). The data should appear as below:

 

Once you have the data set up, go to the "Analyze" option on on the menu bar and then select "Fit Model" from the pull down menu. This should open a dialog box for "Model", shown below:

If you did not assign the variables properly as shown above, this is your second chance. Here you can add Y variables and remove and add effects to your model by hilighting the variable and either clicking the ":remove" button or the "Y" button. When you have it as you like it, click "Run Model". This gives you a "Model Fit" window reporting some of the results of ANOVA, shown below:

For this particular Paradigm, we are really dealing with a repeat measures paradigm. To do this, in the top of the "Model Fit" window, notice the box outlined in blue. Click where it says "Click Here", and from the pull down menu that appears, choose "Repeated Measures". Yet another dialog box opens (shown below).

This enables us to label the levels of our within subjects factor. Since this paradigm deals with a measurement before and after a treatment on the same subject, I used the default name of "Time". When you have done this Click "OK". The results appear at the bottom of the "Model Fit" window.

See results. 

Chi-Square

In order to perform the Chi-Square test of independence in JMP, it is easiest if we have the data in a particular format, that basically consists of categories and counts for those categories. The data window is shown below. Since you will only be dealing with four rows and three columns, it would probably be just as easy to type the data in the data window of JMP so it looks like the following:

Make sure there is an "F" in the upper right box in the frequency column. This tells JMP we are dealing with counts of data. Next go to the "Analyze" option in the file menu and choose "Fit Y by X" from the pull down menu. This opens the "Fit Y by X" dialog box (shown below).

When you first open the window, you will only have variables in the "Columns from [dataset]" box and in the box to the right of the "> Freq >" button. Choose one categorical variable for "X" and the other for "Y" by selecting that variable on the left and clicking the "> Y >" or the "> X >" button, whichever is appropriate. Then click "OK". This will give you the "Y by X" window (shown below)

To get other relevant information, such as the row and column percentages, and other options, click the arrow box beside "Crosstabs". This gives you a pop up menu, from which you can choose percentage values that you may want to report in an analysis. On the JMP analysis page, we show the count, row, column, and total percentages.

See results

 

ANCOVA

To perform an ANCOVA analysis, we need to define our variables properly. We want score1 to be our covariate, active to be our main effect or factor, and score2 to be our dependent variable. To do this, we need to define score1 and score2 as continuous variables, which you do by clicking the box at the top left of each column, and then choose "c Continuous" from the pull down menu. Next define active to be a nominal variable. Do this by choosing "n Nominal" from the pull down menu as described above. Next, define score1 and active to be dependent variables by selecting "X" from the pull down menu in the box at the top right of each column. Do the same for score2, except be sure to select "Y" from the pull down menu. The tops of the columns should appear as below.

Next go to the "Analyze" option on the menu bar at the top of the screen. Then select "Fit Model" from the pull down menu. This gives you the "Model" dialog box. Since we also want to examine the interaction between Active and Score1, we need to define the interaction term. Do this by selecting "active" in the variable list on the left. Then click the " > Add >" button. Next select the bottom "active" in the "Effects in Model" box, and then select "score1" on the left so that active is highlighted in the "Effects in Model" box and "score1" is highlighted in the variable list on the left. Then click the " > Cross > " button (shown below).

This gives you your interaction term.

Then click "Run Model".

 See results

Now we performed the test for a significant interaction. If the interaction is significant we do not want to perform an ANCOVA analysis because we have nothing to gain. Since it is not significant, we proceed with the simple main effect model ANCOVA. Go back to "Analysis" on the menu bar and then choose "Fit Model" again, and you should see the now familiar "Model" dialog box. Highlight the interaction term and then click the "Remove" button (shown below).

The interaction term should have disappeared, so you only have "score1" and "active" in the "Effects in Model" box (shown below).

Finally, click "Run Model".

See results