Variables setup

So far this sheet is behaving much like Excel in what we can enter. However unlike Excel, Jamovi allows us to enter more information about what is contained in each of our variables (the spreadsheet columns). This is known as metadata. To edit this, click Data, then Setup (the icon that looks like a gear on top of a mini-spreadsheet). You will get an extra menu open at the top of the Jamovi screen:

You can see the name of the variable we are editing at the top, in large text. This is the variable Label. Click on any cell in another column to change to that variable - see how the text at the top of the variable setup menu changes.

Directly underneath the large Label area is another text box - this is the Description area, where you can add a little extra text to make it clearer what is contained in our variable. Click on the PartID column and notice how it says ‘Participant ID’ now - this is the full name of the variable, rather than the shorter ‘PartID’ label.

I now want you to change a couple of variable properties:

• You can see that most of our variables have a Name and Description, but two are missing. For the column Age enter the description ‘Age (years)’ and for the Gender column enter the description ‘Gender’.

Variable data types

Underneath the description are some more options. Click on the dropdown menu next to Measure type:

Jamovi has three main data types:

• Nominal
• Ordinal
• Continuous

Nominal and Ordinal are exactly the same as covered in the lecture on data types. ‘Continuous’ is just the combination of Interval and Ratio - Jamovi will work out if the data you enter is Interval or Ratio for you behind-the-scenes based on the numbers you actually have in the data.

Jamovi also has one extra data type, called ‘ID’. This is just used by Jamovi to identify a Participant ID column - it’s exactly the same as a Nominal column, but lets Jamovi know it doesn’t need to worry about labelling levels for it, and to just use whatever is typed into the cells. Don’t worry about the difference too much - just know that you only need to set your PartID column to ‘ID’, and can then forget about it!

Underneath the Measure type options, click the dropdown menu for Data type:

Data types just tell Jamovi what kind of actual data we will be typing in to the cells in the Spreadsheet. These should be fairly obvious:

• Integer: whole numbers without a decimal point (1, 2, 3…)
• Decimal: numbers with a decimal point (1.5, 2.3, 3.0, 4.6…)
• Text: words! (Yes, no, John Smith, football…)

We’ll ignore the ‘Missing values’ section for now.

Coding variables levels

Finally, on the bottom right of the variable viewer is the Levels setup. Here, if we are using categorical data (i.e. Nominal or Ordinal data types), we can set the labels that should appear for each level of that category.

As an example, make sure you’re in the variable setup menu, then click on the Gender column. Notice how the data is entered as the numbers 1 and 2, rather than in text. Change the Measure type to ‘Nominal’, and you should see the numbers 1 and 2 appear in the Levels box.

Click on the number 1 in the Levels box to enter the label for this level. Change this from a ‘1’ to the word ‘Male’. Click on the number 2 and change it to ‘Female’.

Notice how the numbers 1 and 2 move to the lower right under the text you’re typing, so you can still see which label refers to which number. Click anywhere outside the labels box to save these labels.

Notice now that the cells in the Gender column in the spreadsheet change from ‘1’ and ‘2’ to ‘Male’ and ‘Female’. The underlying data is still the same - we’ve just labelled it.

Now go through and check all the variables in our data sheet are setup correctly. You should make sure variables match the following:

• PartID: Participant ID, ID measure type, Integer data type
• Age: Age (years), Continuous, Integer
• Gender: Gender, Nominal, Integer, with 2 levels
• Course: Course Name, Nominal, Integer, with 4 levels
• Weight: Weight (kg), Continuous, Decimal
• Height: Height (cm), Continuous, Decimal
• BIA1: BIA Machine 1 Body Fat (%), Continuous, Decimal
• BIA2: BIA Machine 2 Body Fat (%), Continuous, Decimal
• FitPercept: Fitness Perception, Nominal, Integer, with 5 levels

Once you’ve checked all these variables, click on the circle icon with an Up arrow inside it in the top right of the variable setup menu to hide the menu and return to the main view.

Make sure you save your data regularly - click the three white lines in the top left to access the File menu again, then click Save.

Computing a new variable

We are still missing the BMI variable, so we will create one. You could just create all the data in Excel then copy and paste it into Jamovi, but Jamovi has a Compute Variable function which can do the calculation for us.

Click on the FitPercept column, then click Data, then Compute (the icon that looks like a calculator). This will add a new column after the one our cursor is on (which is why we clicked on FitPercept first!) and open the Computed Variable menu at the top, similar to the variable setup menu:

Jamovi automatically gives our variable a letter of the alphabet as a name - ‘J’ in this case. Change the new variable name to BMI, and then enter Body Mass Index as the description.

Below the name and description is the formula box. Here, we enter a formula to tell Jamovi what to calculate for us - very similar to when we typed functions into Excel, but instead of typing them into one cell and copying and pasting the formula in all the other cells, in Jamovi we just enter the function at the top and it automatically fills it in for the entire column for us.

We can type directly into the formula box, or have Jamovi fill in some parts for us. Click the fx button to see all the options we can enter in the formula box:

Jamovi tells us what each function does if we click on it once - see here I’ve clicked on the MEAN function, and Jamovi gives a short description underneath. Next to the functions are our variables. Double click on any of these to insert them into the function box.

Now, type or copy and paste the following into the function box:

Weight / ((Height/100) * (Height / 100))

Press enter and you should see the BMI column automatically fill in with our BMI values. Notice that any cases (rows) that are missing any Weight or Height data are also left blank in the new BMI column.

Let’s recap: we have looked at the Spreadsheet, which contains our data (cells) arranged as Variables (columns) and Cases (rows). We have looked at the Variable Setup menu to look at the metadata information about our variables and the values they contain. These are the two main components of what we would call data.

The variables and the data are all saved in one Jamovi data file (.omv). Make sure you save your data file as you have made changes to both variables and cases. We will now move on to examine some of our data.

Frequency tables

Let’s also see how many students are on each of the four courses offered. Move BIA1 and BIA2 out of the variables selected, then move Course into the variables selection box.

Click the Frequency tables checkbox underneath our variables selection to add a frequency table to our Results viewer (this will only work for Nominal or Ordinal data types). If you want to make the Results viewer a bit tidier, untick all the other Statistics options, but leave in the N and Missing options:

Here you are presented with frequency as a count, as an overall percentage, and as a cumulative percentage. Top table also tells us about the total number of values and how many are missing values - that is, any cases we have that do not have a valid Course ID.

Notice that our counts in the Frequency table add up to the number of valid Course IDs we have in our data. You need to make sure you always check for any missing values in your data - in this case, you might need to go back and find that missing Course ID, or else exclude the case from any further analyses if Course ID is required.

Histogram

While we expect you to generate graphs in Excel, and produce tables in MS-Word, you can still generate graphs in Jamovi for a quick view of data before formatting them in Excel. Remember, for any coursework, it is generally best to create your final graphs in Excel so you can completely customise how they look.

We will look at two graphs:

• Histogram
• Box plot

We will continue to use our Body Composition.omv data set. Still in the Descriptive menu, clear all the variables from our variable selection, and untick the Frequency table option. Move Weight into the variables selection and turn back on the Default descriptive statistics - tick the boxes for N, Missing, Mean, Median, Std. Deviation, Minimum, and Maximum.

Then, you can click the Statistic menu bar again to fold it out of the way, and click on the Plots menu bar. Click on Histogram to generate a simple histogram in our Results viewer.

This is a histogram for our entire data set, but we can split this by Gender exactly as we did for the Descriptives. Move Gender into the Split By box and see what happens to our histogram:

You can split your data using any Nominal or Ordinal variable - try moving other variables in and out of the Split By box to see how the histogram changes.

Box plots

Boxplots may also be useful to visually inspect your data, especially if you are looking for outliers. We will use a box plot to check for any outliers in our height data. Clear the selection boxes, then move Height into the variables selection, and Split By Gender again. Then, click the Box plot tick box to show a box plot in the Results viewer:

Remember in a boxplot:

• Top whisker = 1.5 x interquartile range
• Top of box = 3rd quartile
• Line in the middle = median
• Bottom of box = 1st quartile
• Bottom whisker = 1.5 x interquartile range

Any outliers are shown with circles beyond the whiskers of the box plot. There are no recognised outliers in our Height data.

Feel free to click around now on any other Plot or Statistics options you think look interesting, and try exploring different variables and different ways of splitting the data. Remember, if you want to remove something from the Results viewer, just untick it.

Tests of normality (normal distribution)

While we could simply look at a histogram to visually inspect for a normal distribution, there are a number of statistical tests we can also perform. The main statistical test for normality in Jamovi is the Shapiro-Wilk (SW) test.

Let us examine the distribution of male and female height data. To compare the two sets of independent data we need to make sure that both male and female data do not deviate significantly from a normal distribution. (Note: testing repeated measures data requires a slightly different test, but here we will examine normal distribution for independent samples and look at repeated measures later)

Still in the Descriptives menu, move Height into the variables selection, and split the data by Gender. Then, under the Statistics menu, click the Shapiro-Wilk tick box:

This will add rows for Shapiro-Wilk W and Shapiro-Wilk p values to our descriptives table in the Results viewer.

Here we are testing whether the distribution of Height (by gender) is normally distributed so we can use parametric tests later. We want to see whether there is a significant deviation from normality, i.e. whether the p value result is less than 0.05 for Height for either Male or Female groups. Both groups have p > 0.05. This means there is no significant deviation from a normal distribution, so the test of normality has been satisfied.

Now you try: Perform a test of normality for Weight by Gender using the Body Composition.omv data set.

A simple SW test is ok to start, but you should still perform further checks on your data. The first two we will examine are Skewness and Kurtosis.

Still in the Descriptives menu, choose the Height variable split by Gender. Under Statistics, tick Skewness and Kurtosis.

The first way we can examine this data is to look at the absolute values for Skewness and Kurtosis.

• Male skewness = 1.028
• Male kurtosis = 0.959
• Female skewness = -0.036
• Female kurtosis = -0.481

A rule of thumb suggests that -0.8 to +0.8 is acceptable for skewness and -3 to +3 is acceptable for kurtosis. Note that the skewness for males falls outside the -0.8 to 0.8, but all other values are well within their ranges. Strictly speaking we could suggest that the male height data is not normally distributed, and has a slight positive skew (i.e. a pile up of data on the left hand side).

Turn back on the box plots for this data. You can see the slight positive skew for the male data with a longer positive tail:

A further way of examining this data is to convert the height skewness and kurtosis to a ‘standardised’ value. We do this by taking the value and dividing it by the standard error of its value. You can find the standard error values in the Descriptives table just underneath the Skewness and Kurtosis.

Before we work out our standardised values though, let’s change the number of decimal places Jamovi is using to show our Results, so we can be more accurate. In the far top right-hand corner of Jamovi is three white dots - click these to open the Preferences panel. Here, change the Number format dropdown menu to 3 dp (3 decimal places). Click anywhere outside the Preferences panel to close it again.

Now we can calculate our standardised skewness and kurtosis values:

These values can now be compared to the values we would expect to get by chance alone, so a value less than -1.96 or greater than 1.96 would suggest that this distribution is significantly skewed or shows significant kurtosis. Both of our values are within this range so we can be happy that our data satisfies the condition of normality.

Even though we would like to think that statistics should be clear-cut in giving us answers, you can now see that it can often requires careful interpretation of your results. Here we have seen two tests that show our data is normally distributed, with one test that shows it is not! This is where your experience of working with numbers is going to be very useful.

Homogeneity of variance

This is a check to see that the samples we are testing come from populations with a similar variance.

As with the normality test, there is a statistical test and a rule of thumb. Starting with the rule of thumb, we want to make sure that the two sample variances differ by a factor of less than 3.

Imagine we wished to compare male and female height - we therefore need to compare the sample variance of the male height and the female height.

We can view these variances in Jamovi - still with Height split by Gender selected, tick the variance check box. This will display the variances for Male and Female groups in our Descriptives table in the Results viewer.

To test our rule of thumb, take whichever of the two is the larger variance and divide it by the smaller variance. In our case, that would be:

65.368 / 55.649 = 1.17

1.17 is well within our rule of thumb of 3, so we can assume homogeneity of variance. But, let’s also test it properly. The statistical test we will use for this is Levene’s Test of Homogeneity.

To do this, we actually need to go to a new menu option. Click the right-arrow in the circle at the top right of the Descriptives menu to close it. Then, still in the Analyses menu, click the T-Tests icon, and choose Independent Samples T-Test from the dropdown menu.

This will open the T-Test menu. You’ll see that it looks very similar to the Descriptives menu - we choose our variables of interest and how we want to split them (here called Grouping Variable) at the top, and just tick the things we want Jamovi to show us underneath.

Move Height into our variables selection and split by Gender again. Don’t worry about any of the other options, but just tick the Homogeneity test option under Assumptions.

Notice how this gives us a new area in the Results viewer, under the heading Independent Samples T-Test. But, our Descriptives table is still displayed above it (you might need to scroll back up). If you click on the descriptives table in the Results, the descriptive menu will open back up in the middle, and you can make any changes. This is how your results are stored and edited in Jamovi - just click them in the viewer to open the menu for that particular result. If you want to remove a result from your viewer entirely, right-click on it’s title and select ‘Remove’ from the pop-up menu.

We won’t worry about the T-Test results for now - just look at the table titled Homogeneity of Variances Test (Levene’s). Here, we are looking to see if the test result is significant, just as we did for the normality test. The non-significant value of p = 0.675 says that the variances are not significantly different to each other and we can continue with our parametric tests.

Now you try: Repeat the test of Homogeneity of Variance for Weight by Gender.