Linear mixed model

Overview

This function uses fitlme to fit one-way or two-way linear mixed models to data with with an optional grouping variable.

This is a useful approach when the data exhibit nesting. For example, the data include values measured from several specimens from the same person.

Models can be of the form (using Wilkinson notation):

• y ~ factor_1
• y ~ factor_1 + (1 | grouping)
• y ~ factor_1 + factor_2 + (factor_1 * factor_2)
• y ~ factor_1 + factor_2 + (factor_1 * factor_2) + (1 | grouping)

Main effects are dervied using anova

Post-hoc tests are calculated using coefTest and then corrected using the Holm-Bonferroni method.

Examples

• Linear mixed model

Function arguments

function stats = linear_mixed_model(t, data_label, f1_label, options);

    arguments
        t (:,:) table
        data_label (1,1) string
        f1_label (1,1) string
        options.f2_label (1,1) string = ""
        options.grouping_label (1,1) string = ""
        options.figure_handle (1,1) double = 0
        options.subplot_handle (1,1) double = 0

    end

Details

Inputs

Parameter	Values	Needed	Description
t	table	Required	Data in tidyverse format
data_label	string	Required	Table variable for test data
f1_label	string	Required	Table variable for factor 1
f2_label	string	Optional	Table variable for factor 2
grouping_label	string	Optional	Table variable for grouping parameter
figure_handle	double	Optional	Figure handle, if provided code will create a new figure with this handle showing data as a simple swarmchart
subplot_handle	double	Optional	Subplot handle, if provided code will add a simple swarmchart to the subplot

Return values

Parameter	Fields	Description
stats		Structure
	main_effects	A main effects table, derived from anova
	post_hoc	A table showing post-hoc tests corrected for multiple comparisons using the Holm-Bonferroni method
	model_string	A string with the model that was tested in Wilkinson notation
	main_effects_string	A string summarizing the main effects