Conducting a Standard Multiple Regression in jamovi

Star Wars is a much beloved franchise. Some characters and plots however are more
beloved than others by the fan base. A researcher decided to conduct a quick social media
poll on the level of agreement with a number of Star Wars opinions known to have varying
levels of acceptance or incredulity to die-hard fans. She was particularly interested in
predictors of agreement with the opinion that JarJar Binks is awesome as this opinion seems
to be the source of great division in the fandom. A total of 100 survey respondents rated
their agreement with four beliefs on a 5-point scale where 1 is strongly disagree and 5 is
strongly agree.

The Star Wars opinions rated were:

• JarJar Binks is awesome.
• The Kylo Ren and Rey romance makes sense.
• Han and Chewie rock.
• Boba Fett is over-rated.

Step 1 – Taking a look at the data.

Our four variables have been

specified as ordinal variables in
Measure type. The anchor points of
the likert scale (1 = strongly
disagree, 5 = strongly agree) have
also been entered

In the data spreadsheet are four

columns of data representing the
agreement ratings given to each of
the four Star Wars opinions. Each
row represents a person who has
Step 2 – Navigating to the Linear Regression menu. rated each of the four opinions.

On the Analyses tab select

the Regression menu, then
select Linear Regression.

v.4 – 10 July 2021 1

 Conducting a Standard Multiple Regression in jamovi

Step 3 – Selecting analysis options

The first thing we need to do is tell jamovi what variable is our criterion/dependent variable
and which are our predictors. It is important to note that jamovi refers to continuous
predictors as covariates and categorical predictors as factors. In our case all our predictors
are continuous so we will need to move them across to the covariates box.

The default output we get from specifying for the variables for the analysis appears below.

We are given the R and R2 for the

whole model as well as
unstandardised regression
coefficients (labelled Estimate) and
associated significance testing.

However, there are a few more

elements and tests we require for
completeness.

From the drop down lists we’ll ask

for options from Model Fit and
Model Coefficients.

v.4 – 10 July 2021 2

 Conducting a Standard Multiple Regression in jamovi

Under Model Fit, in addition to the R and R2 already set as a default, we’ll ask for adjusted
R2 and the Overall Model F test.

Our Model Fit Measures table

now includes the omnibus F test
for the full regression model,
and the adjusted R2.

Under the Model Coefficients drop down menu, we’ll ask for confidence intervals for our
unstandardised regression coefficients, Bs, (under Estimate) as well as asking for
standardised regression coefficients, betas, (under Standardised Estimate) and their
confidence intervals as well. You don’t need to report both sets of confidence intervals but
have done so here to demonstrate what the output looks like for both.

The Model Coefficients table has now substantially expanded to include standardised and
unstandardised results.

v.4 – 10 July 2021 3

 Conducting a Standard Multiple Regression in jamovi

One key statistic that is conventionally reported with standard multiple regression results,
namely squared semi-partial correlations, or r2a(b,c), are not available via the Linear
Regression menu (as at jamovi version 1.8.4). To obtain these we need to run some
separate analysis.

Head to the Analyses ribbon, select Regression and then Partial Correlation

In the current jamovi set up for

obtaining partial and semi-partial
correlations they need to be specified
one at a time. In other words we need
to specify a pair of variables to be
correlated and which variables to use
as control variables, or variables that
will be partialled out of the
calculations.

Before we start let’s make sure the options we need have been selected. The most
important selection is to change the selection to “Semipartial” under Correlation Type as
Partial is the default. Also untick “Report significance” under Additional Options to keep the
output minimal and easy to interpret.

v.4 – 10 July 2021 4

 Conducting a Standard Multiple Regression in jamovi

There are three semi-partial correlations we need to request. These are between our
criterion/dependent variable JarJar Binks is awesome with each of the three predictor
variables/opinions, controlling for the respective remaining two each time. Below is how
you will specify each semi-partial correlation that you need for this regression model.

Note that after running each semi-partial correlation you will need to click underneath the
output so that jamovi knows to create an additional output, rather than editing the one you
just created. Jamovi will edit whichever output is currently highlighted in white. To ensure
you get additional output rather than overwritten output ensure you have clicked
underneath the output created and it has turned grey as can be seen on the right hand side
below.

Below is the first semi-partial correlation.

vs

v.4 – 10 July 2021 5

 Conducting a Standard Multiple Regression in jamovi

Our three semi-partial correlation runs will provide the following output:

In each of these semi-

partial correlation tables
the coefficient we want is
listed in the column
heading with the
predictor name of
interest as circled on
each.

The statistic we need to report in our regression write up is the SQUARED

semi-partial correlation. So each of these three semi-partial correlations need
to be squared manually.

v.4 – 10 July 2021 6

 Conducting a Standard Multiple Regression in jamovi

Step 4 – Finding the components for reporting.

There are two parts to our standard multiple regression write up. The overall model results
and then results pertaining to the role of each individual predictor in the model.

The components we need for the overall model results are:

1. The F statistic, dfs and p value – the omnibus ANOVA result for the full model
2. Effect sizes in the form of R2 and adjusted R2.

The Write Up (Part 1):

A standard multiple regression was conducted to determine the extent to which three
opinions in relation to Star Wars characters and relationships predict level of agreement
that JarJar Binks is awesome. The three opinions combined explained 23.8% of the variance
in agreement with the JarJar Binks opinion, representing a significant proportion of variance
explained, adj. R2 = .21, F(3,96) = 9.99, p <.001.

v.4 – 10 July 2021 7

 Conducting a Standard Multiple Regression in jamovi

Getting the parts for the Write Up (Part 2):

We’ll create a table to contain the results about the individual predictors. This will require
us to consolidate the information from the Model Coefficients table from the regression
output as well as the semi-partial correlations we ran separately. We’ll put the regression
output elements in the table first.

The elements needed for the regression table are:

1. p values for each predictor – to determine the significance of each predictor’s

contribution to the model
2. Effect sizes in the form of Bs, 𝛽s and r2a(b,c) (Squared semi-partial correlations).
3. Confidence intervals – around the standardised regression coefficients (𝛽).

Table 1

Standard Multiple Regression results for the prediction of agreement that JarJar Binks is
awesome from three additional opinions of Star Wars characters and plotlines

𝜷 95% CI for 𝜷 B p r2a(b,c)

LL UL
Kylo and Ren romance .39 .20 .58 0.34 <.001
makes sense
Han and Chewie rock -.20 -.38 -.01 -0.19 .039
Boba Fett is over-rated .07 -.13 .27 0.06 .475

v.4 – 10 July 2021 8

 Conducting a Standard Multiple Regression in jamovi

Getting our r2a(b,c) from our semi-partial correlation output:

Square this value for the r2a(b,c) for

the Kylo & Rey romance makes sense
predictor.

0.358962 = 0.129

Square this value for the r2a(b,c) for

the Han and Chewie rock predictor.

-0.186712 = 0.035

Square this value for the r2a(b,c) for

the Boba Fett is over-rated predictor.

0.063912 = 0.004

We can now complete our table below:

Table 1

Standard Multiple Regression results for the prediction of agreement that JarJar Binks is
awesome from three additional opinions of Star Wars characters and plotlines

𝜷 95% CI for 𝜷 B p r2a(b,c)

LL UL
Kylo and Ren romance .39 .20 .58 0.34 <.001 .129
makes sense
Han and Chewie rock -.20 -.38 -.01 -0.19 .039 .035
Boba Fett is over-rated .07 -.13 .27 0.06 .475 .004

v.4 – 10 July 2021 9

 Conducting a Standard Multiple Regression in jamovi

The Write Up (Part 2):

The results for each individual predictor within the regression model are shown in Table 1.
Increased agreement with the opinion that the Kylo Ren and Rey romance makes sense was
significantly associated with increased agreement that JarJar Binks is awesome, while
increased agreement that Han and Chewie rock was significantly associated with decreased
agreement regarding JarJar Binks. Opinions regarding Boba Fett did not add a significant
unique contribution to the regression model.

Table 1

Standard Multiple Regression results for the prediction of agreement that JarJar Binks is
awesome from three additional opinions of Star Wars characters and plotlines

𝜷 95% CI for 𝜷 B p r2a(b,c)

LL UL
Kylo and Ren romance .39 .20 .58 0.34 <.001 .129
makes sense
Han and Chewie rock -.20 -.38 -.01 -0.19 .039 .035
Boba Fett is over-rated .07 -.13 .27 0.06 .475 .004

Created by Janine Lurie in consultation with the Statistics Working Group within the School of Psychology,
University of Queensland 1

Based on jamovi v.1.8.4 2

1
The Statistics Working Group was formed in November 2020 to review the use of statistical packages in
teaching across the core undergraduate statistics unit. The working group is led by Winnifred Louis and Philip
Grove, with contributions from Timothy Ballard, Stefanie Becker, Jo Brown, Jenny Burt, Nathan Evans, Mark
Horswill, David Sewell, Eric Vanman, Bill von Hippel, Courtney von Hippel, Zoe Walter, and Brendan Zietsch.
2
The jamovi project (2021). jamovi (Version 1.8.4) [Computer Software]. Retrieved from
https://www.jamovi.org

v.4 – 10 July 2021 10