Multinomial Logistic Regression for Categorically-Coded Verbal Data
Instructions:
Step 1: Choose an appropriate model:
A
One Factor Model
fits a Bayesian multinomial model to your data. It specifies a
random effect for the identifier and includes only Factor1.
A
Two Factor Model with Interaction
fits a Bayesian multinomial model to your data. It specifies a
random effect for the identifier and includes an interaction between the two factors.
A
Two Factor Model without Interaction
fits the same two factor model, but without
the interaction effect.
Step 2: Format your data file as follows:
a. Label the first predictive factor as
Factor1.
b. If you have a second predictive factor or dimension, label it as
Factor2.
c. Label the source or identifier for the data as
Case.
d. Label your outcome variable, your coding, as
Dimension.
e. Delete any columns holding actual verbal data.
f. Save the copy as a CSV (comma separated values) format.
Step 3: Load your CSV file using the Browse button on the left below.
Step 4: Check the data by comparing the appropriate data table with your own frequency table.
Choose One Factor Data Table for a one factor model and Two Factor Data Table for a two factor model.
If there are discrepancies, go back to your data file to make sure the data are correct and labelled appropriately. This is
your chance to correct any misspellings or incorrect inputs. Please be sure that your coding dimension column header is
Dimension, your factor column headers are Factor1 and Factor2, and your identifier column header is Case. All of these headings
are case sensitive.
Step 4: Run the regression by clicking on the tab for the appropriate model.
Keep in mind that some models can take minutes to run. Do not refresh the page.
Step 5: Read the results.
The bottom table shows the coefficients for the model and indicates the statistical significance of each term, starting with
the intercept and moving through the factors and their interactions. Post mean is the posterior mean, which is the point estimate
for the coefficient. If your effective sample size is much smaller than the sample size, please be cautious in using your estimates.
More about this app can be found in Coding Streams of Language, Chapter 9 downloadable at
https://wac.colostate.edu/books/practice/codingstreams/