farmer survey UK
Cynthia Naydani
2026-01-15
Getting Started
Install packages
Import data
df_raw <- read_excel("FarmerSurvey_Master_260123.xlsx",
sheet = "Raw") %>%
clean_names() # Uses janitor packages to clean names of objects
view (df_raw)Data Preparation
Renaming columns
Extract column names into a dataframe
var_names <- tibble(old_name = names(df_raw))
view(var_names)Create a renaming table
# Naming conventions:
# One Health theme: wb = wellbeing, aw = animal welfare, pl = planet
# COM-B: cap = capability, mot = motivation, opp = opportunity
rename_key <- tribble(
~old_name, ~new_name, ~theme, ~com_b, ~reverse_scored,
"id", "id", NA, NA, NA,
"start_time", "start_time", NA, NA, NA,
"what_are_the_last_four_digits_of_your_phone_number_this_cannot_be_used_to_identify_you_we_ask_this_only_so_that_you_can_choose_to_withdraw_from_the_study_within_seven_7_days_of_submission_or", "phone", NA, NA, NA,
"what_country_do_you_live_in", "country", NA, NA, NA,
"what_is_your_age_in_years_this_question_is_optional", "age", NA, NA, NA,
"what_is_your_gender_this_question_is_optional", "gender", NA, NA, NA,
"how_long_have_you_been_farming_if_you_were_raised_in_farming_please_include_your_childhood_years_as_applicable", "farming_time", NA, NA, NA,
"what_is_your_farming_situation", "farm_type", NA, NA, NA,
"what_type_s_of_livestock_do_you_farm_please_select_all_that_apply", "livestock", NA, NA, NA,
"do_you_also_have_arable_crops", "arable", NA, NA, NA,
"is_your_farm_fully_organic", "organic", NA, NA, NA,
"do_you_consider_your_farm_to_be_regenerative_agroecological_if_you_are_unsure_please_select_no", "regen", NA, NA, NA,
"is_your_farm_a_member_of_any_assurance_schemes", "schemes", NA, NA, NA,
"which_one_s", "schemes_which", NA, NA, NA,
"is_there_anything_else_you_would_like_us_to_know_about_your_farm_farming_system_please_do_not_disclose_any_personal_information_like_farm_name_or_specific_location_this_question_is_optional", "farm_freetext", NA, NA, NA,
"i_am_self_confident", "wb_mot_confident", "wellbeing", "motivation", FALSE,
"other_farmers_come_to_me_for_advice", "wb_mot_advice", "wellbeing", "motivation", FALSE,
"i_am_physically_healthy", "wb_cap_healthy", "wellbeing", "capability", FALSE,
"overall_i_am_satisfied_with_myself", "wb_mot_satisfied", "wellbeing", "motivation", FALSE,
"when_i_learn_new_information_i_put_it_into_action", "wb_mot_action", "wellbeing", "motivation", FALSE,
"i_enjoy_interacting_with_other_people", "wb_mot_interact", "wellbeing", "motivation", FALSE,
"i_have_broken_bad_habits_to_improve_my_wellbeing", "wb_cap_badhabit", "wellbeing", "capability", FALSE,
"i_have_access_to_the_resources_i_need_such_as_time_and_money_to_live_a_good_life", "wb_opp_money", "wellbeing", "opportunity", FALSE,
"it_is_important_to_me_to_have_the_approval_of_my_friends_and_family", "wb_opp_approval", "wellbeing", "opportunity", FALSE,
"i_adapt_easily_to_change", "wb_mot_adapt", "wellbeing", "motivation", FALSE,
"i_find_looking_after_my_own_well_being_rewarding", "wb_mot_reward", "wellbeing", "motivation", FALSE,
"i_can_do_things_as_well_as_most_other_people", "wb_mot_ability", "wellbeing", "motivation", FALSE,
"i_have_people_who_i_can_turn_to_for_help", "wb_opp_help", "wellbeing", "opportunity", FALSE,
"i_take_a_positive_attitude_toward_myself", "wb_mot_posatt", "wellbeing", "motivation", FALSE,
"i_know_ways_to_improve_my_own_mental_wellbeing", "wb_cap_improve", "wellbeing", "capability", FALSE,
"i_have_people_around_me_who_support_me", "wb_opp_support", "wellbeing", "opportunity", FALSE,
"is_there_anything_else_you_would_like_us_to_know_about_your_wellbeing_please_do_not_disclose_any_personal_information_that_could_identify_you_this_question_is_optional", "wellbeing_freetext", NA, NA, NA,
"i_know_what_to_do_to_improve_the_wellbeing_of_my_animals", "aw_cap_improve", "animal", "capability", FALSE,
"i_have_access_to_all_the_resources_e_g_time_money_i_need_to_take_good_care_of_my_animals", "aw_opp_resource", "animal", "opportunity", FALSE,
"improving_the_wellbeing_of_my_animals_will_be_beneficial_for_my_farm_and_business", "aw_mot_benefit", "animal", "motivation", FALSE,
"the_benefits_of_taking_good_care_of_my_animals_outweighs_the_costs", "aw_mot_outweigh", "animal", "motivation", FALSE,
"i_feel_emotionally_connected_to_my_animals", "aw_mot_connect", "animal", "motivation", FALSE,
"other_farmers_i_know_take_steps_to_improve_the_wellbeing_of_their_animals", "aw_opp_farmers", "animal", "opportunity", FALSE,
"my_physical_health_makes_it_difficult_to_take_good_care_of_my_animals", "aw_cap_health", "animal", "capability", TRUE,
"a_main_responsibility_of_a_farmer_is_to_uphold_the_wellbeing_of_his_her_animals", "aw_mot_resp", "animal", "motivation", FALSE,
"the_wellbeing_of_my_animals_is_one_of_the_most_important_things_in_my_life", "aw_mot_priority", "animal", "motivation", FALSE,
"taking_good_care_of_my_animals_is_within_my_control", "aw_mot_control", "animal", "motivation", FALSE,
"others_i_work_with_e_g_purchasers_industry_associates_pressure_me_to_improve_the_wellbeing_of_my_animals", "aw_opp_pressure", "animal", "opportunity", FALSE,
"i_know_my_animals_have_feelings", "aw_cap_feelings", "animal", "capability", FALSE,
"i_have_other_people_who_help_me_improve_the_lives_of_my_animals", "aw_opp_others", "animal", "opportunity", FALSE,
"i_care_about_how_my_animals_feel", "aw_mot_feel", "animal", "motivation", FALSE,
"i_will_experience_benefits_rewards_if_i_take_steps_to_improve_the_wellbeing_of_my_animals", "aw_mot_worth", "animal", "motivation", FALSE,
"if_animal_welfare_is_maximised_there_are_negative_impacts_on_farm_productivity_and_economics", "aw_mot_tradeoff", "animal", "motivation", TRUE,
"is_there_anything_else_you_would_like_us_to_know_about_your_views_and_experiences_with_your_livestock_this_question_is_optional", "livestock_freetext", NA, NA, NA,
"if_i_do_things_to_reduce_my_environmental_impact_the_health_and_wellbeing_of_my_animals_will_also_improve", "pl_mot_mutual", "planet", "motivation", FALSE,
"i_am_worried_about_the_health_of_the_planet", "pl_mot_ecoanxiety", "planet", "motivation", FALSE,
"i_feel_pressure_from_others_to_reduce_my_environmental_impact", "pl_opp_pressure", "planet", "opportunity", FALSE,
"i_have_access_to_all_the_resources_i_need_e_g_time_money_to_take_actions_that_minimise_negative_environmental_impacts", "pl_opp_resource", "planet", "opportunity", FALSE,
"the_environment_is_being_harmed_by_the_combined_actions_of_individual_people", "pl_cap_harm", "planet", "capability", FALSE,
"my_poor_physical_health_makes_me_unable_to_do_things_that_would_benefit_the_environment_e_g_walking_places_instead_of_driving", "pl_cap_health", "planet", "capability", TRUE,
"i_know_what_i_have_to_do_to_be_more_environmentally_friendly", "pl_cap_know", "planet", "capability", FALSE,
"caring_for_the_environment_is_our_social_responsibility", "pl_opp_know", "planet", "opportunity", FALSE,
"i_feel_a_sense_of_connection_to_the_planet_and_to_wildlife", "pl_mot_connect", "planet", "motivation", FALSE,
"taking_care_of_the_environment_comes_at_the_expense_of_my_farm_efficiency_productivity", "pl_mot_productivity", "planet", "motivation", TRUE,
"a_consideration_for_the_environment_is_an_important_part_of_being_a_farmer", "pl_mot_identity", "planet", "motivation", FALSE,
"it_is_within_my_power_to_positively_impact_the_environment", "pl_mot_power", "planet", "motivation", FALSE,
"the_benefits_of_preserving_the_environment_outweigh_the_downsides_e_g_costs", "pl_mot_outweigh", "planet", "motivation", FALSE,
"other_farmers_i_know_are_taking_steps_to_minimise_negative_environmental_impacts", "pl_opp_peers", "planet", "opportunity", FALSE,
"environmental_conservation_is_among_my_top_goals", "pl_mot_peers", "planet", "motivation", FALSE,
"i_will_experience_benefits_rewards_if_i_take_action_to_preserve_the_environment", "pl_mot_benfit", "planet", "motivation", FALSE,
"is_there_anything_else_you_would_like_us_to_know_about_your_views_on_the_environment_this_question_is_optional", "planet_freetext", NA, NA, NA,
"is_there_anything_else_you_would_like_to_add_this_question_is_optional", "anything_else", NA, NA, NA,
)Safety Check
# Check for unmatched column names with
setdiff(rename_key$old_name, names(df_raw))## character(0)# Check for duplicated new names in the rename key
rename_key %>%
count(new_name) %>%
filter(n > 1)## # A tibble: 0 × 2
## # ℹ 2 variables: new_name <chr>, n <int>Apply the renaming table to the dataframe
# Create a new data frame (df_tidy) to rename columns in df_raw using a lookup table (rename_key)
df_tidy <- df_raw %>%
rename_with( # For each selected column name (.x),
# find its position in rename_key$old_name
# and replace it with the corresponding new_name
~ rename_key$new_name[
match(.x, rename_key$old_name)
],
# Apply this renaming only to columns listed in rename_key$old_name
.cols = rename_key$old_name
)
# save the renaming key as a .csv file for reproducibilty
write_csv(rename_key, "variable_rename_key.csv")
# Now we have a new dataframe called 'df_tidy' which is the same as df_raw but with the renamed column headings to use in subsequent steps of the analyses. Code Likert Data
Define Likert mapping
# This identifies the five levels of the likert scale used in the original dataset
likert_levels_5 <- c(
"Strongly disagree",
"Disagree",
"Neutral",
"Agree",
"Strongly agree"
)Identify likert items
This is possible because in an earlier chunk, metadata were assigned to the statement columns with info regarding their Theme, COM-B component, and if they were reverse coded
likert_items <- rename_key %>%
filter(!is.na(theme), !is.na(com_b)) %>% # keeps only scored items
pull(new_name) %>%
unique()
view(likert_items) # This gives us 48 items in the likert_items string which is correct as there are 16*3 likert statementsConvert text to ordered factor –> numeric
# Convert the Likert-scale columns to ordered factors
df_tidy <- df_tidy %>%
mutate(across(
all_of(likert_items), # selects all of (and only) the likert columns
~ factor(., levels = likert_levels_5, ordered = TRUE)
)) # using the specified 5-point likert levels, turn likert columns into ordered factors (changes the class from character to ordinal)
# Define a numeric mapping for 5-point Likert responses
map_5 <- c(
"Strongly disagree" = 1,
"Disagree" = 2,
"Neutral" = 3,
"Agree" = 4,
"Strongly agree" = 5
)
# Recode likert items from ordered factors to numeric values
df_tidy <- df_tidy %>%
mutate(across(all_of(likert_items), #select all the likert items
~ unname(map_5[as.character(.)]) %>% as.numeric())) #convert factors to characters, map to numeric values using map_5, and coerce to numbers
# NB: we convert factor --> character because converting a factor directly to numeric would corrupt the scale by indexing the map_5 as factor integers rather than labels. The latter (labels) is correct for Likert scales. If we converted to numeric, we could get results that are incorrect and behave unpredictably depending on level ordering. In short, we need text labels to safely map them to predefined numeric scales. If we didn't, we would accidentally use R's internal integer codes instead of intended Likert scoring.
# if labels don't match exactly we will get NAs so let's do a quick check:
na_introduced <- colSums(is.na(df_tidy[ , likert_items]))
na_introduced[na_introduced > 0] # inspect if any appear. ## pl_cap_harm
## 1# one does, as this somehow was missed by the respondent. This is fine as the missing value won't break the workflow since we are using aggregate scores not PCA. The Mann-Whitney tests will automatically drop NAs. Reverse code items
reverse_items <- rename_key %>%
filter(reverse_scored == TRUE)%>%
pull(new_name)# reverse code the identified items in a new dataframe called 'df_reverse'
df_tidy <- df_tidy %>%
mutate(across(all_of(reverse_items), ~ 6 - .))# List of reversed items
identical(df_tidy$aw_cap_health, df_tidy$aw_cap_health)## [1] TRUE# Sanity checks
sapply(df_tidy[ , reverse_items], class) # should be "numeric"## aw_cap_health aw_mot_tradeoff pl_cap_health pl_mot_productivity
## "numeric" "numeric" "numeric" "numeric"summary(df_tidy[ , reverse_items]) # ranges should be 1..5## aw_cap_health aw_mot_tradeoff pl_cap_health pl_mot_productivity
## Min. :2.000 Min. :1.000 Min. :1.000 Min. :1.000
## 1st Qu.:4.000 1st Qu.:3.000 1st Qu.:4.000 1st Qu.:2.250
## Median :4.000 Median :4.000 Median :4.000 Median :3.000
## Mean :3.909 Mean :3.545 Mean :3.682 Mean :2.909
## 3rd Qu.:4.000 3rd Qu.:4.750 3rd Qu.:4.000 3rd Qu.:3.750
## Max. :5.000 Max. :5.000 Max. :5.000 Max. :5.000Create Composite Scores
# Map of items → theme & COM-B (only Likert items)
item_map <- rename_key %>%
filter(!is.na(theme), !is.na(com_b)) %>%
select(item = new_name, theme, com_b) %>%
distinct()
id_cols <- c("id", "regen")
# Long format with metadata
df_long <- df_tidy %>%
pivot_longer(
cols = all_of(item_map$item),
names_to = "item",
values_to = "score"
) %>%
left_join(item_map, by = "item")
#ensure farmer type is a 2-level factor
df_long <- df_long %>% #df long creates a long-format dataset where each row represent one participant, one item, and one score
mutate(regen = factor(regen, levels = c("Yes", "No")))df_long <- df_long %>%
# fix any inconsistent yes/no data
mutate(
regen = trimws(regen),
regen = tools::toTitleCase(tolower(regen))
) %>%
# assign clean labels
mutate(
regen = factor(
regen,
levels = c("Yes", "No"),
labels = c("Regenerative", "Conventional")
)
)Create COM-B x theme composites (both sum and mean)
Because one participant has a missing item, the mean may be preferable (not influenced by the non-missing value). By computing both, mean will be primary and sum as a sensitivity check
df_composites <- df_long %>%
group_by(across(all_of(id_cols)), theme, com_b) %>% #groups data by participant ID (id_cols), theme, and COM-B component.
summarise(
composite_sum = sum(score, na.rm = TRUE), #adds up all item scores contributing to that construct, producing a total scale score = overall magnitude
composite_mean = mean(score, na.rm = TRUE), #average score across items, ignores missing values, standardises scale to 1-5 regardless of number of items. more interpretable for likert scales
n_items = sum(!is.na(score)), # counts how many non-missing items contribute to that composite
.groups = "drop" # removes grouping structure after summarising which can otherwise influence results downstream
) %>%
unite("theme_com_b", theme, com_b, remove = FALSE) # creates new composite variable for plotting, summarising, modellingSanity check
#How many items contributed to each composite (by design)?
df_composites %>%
count(theme, com_b, n_items)## # A tibble: 10 × 4
## theme com_b n_items n
## <chr> <chr> <int> <int>
## 1 animal capability 3 22
## 2 animal motivation 9 22
## 3 animal opportunity 4 22
## 4 planet capability 2 1
## 5 planet capability 3 21
## 6 planet motivation 9 22
## 7 planet opportunity 4 22
## 8 wellbeing capability 3 22
## 9 wellbeing motivation 9 22
## 10 wellbeing opportunity 4 22# Quick distribution snapshot
df_composites %>%
group_by(theme_com_b, regen) %>%
summarise(
n = dplyr::n(),
median_sum = median(composite_sum),
IQR_sum = IQR(composite_sum),
median_mean = median(composite_mean),
IQR_mean = IQR(composite_mean),
.groups = "drop"
)## # A tibble: 18 × 7
## theme_com_b regen n median_sum IQR_sum median_mean IQR_mean
## <chr> <fct> <int> <dbl> <dbl> <dbl> <dbl>
## 1 animal_capability Regenera… 11 13 1 4.33 0.333
## 2 animal_capability Conventi… 11 12 1.5 4 0.5
## 3 animal_motivation Regenera… 11 39 2.5 4.33 0.278
## 4 animal_motivation Conventi… 11 39 4 4.33 0.444
## 5 animal_opportunity Regenera… 11 14 3 3.5 0.75
## 6 animal_opportunity Conventi… 11 15 3.5 3.75 0.875
## 7 planet_capability Regenera… 11 13 2 4.33 0.667
## 8 planet_capability Conventi… 11 11 2 3.67 0.667
## 9 planet_motivation Regenera… 11 36 3.5 4 0.389
## 10 planet_motivation Conventi… 11 37 9.5 4.11 1.06
## 11 planet_opportunity Regenera… 11 14 1.5 3.5 0.375
## 12 planet_opportunity Conventi… 11 14 2.5 3.5 0.625
## 13 wellbeing_capability Regenera… 11 10 1.5 3.33 0.500
## 14 wellbeing_capability Conventi… 11 10 4 3.33 1.33
## 15 wellbeing_motivation Regenera… 11 33 6 3.67 0.667
## 16 wellbeing_motivation Conventi… 11 32 7.5 3.56 0.833
## 17 wellbeing_opportunity Regenera… 11 15 2.5 3.75 0.625
## 18 wellbeing_opportunity Conventi… 11 15 4.5 3.75 1.12Two rows for planet x capability as one respondent skipped one question they are parsed onto a separate row. Overall, the above code transforms item-level Likert data into participant-level construct scores aligned with COM-B x One Health themes, then validating structural integrity and inspecting group-level distributions.
Mann-Whitney U tests for each One Health theme x COM-B composite
library (rstatix)
mw_results <- df_composites %>%
group_by(theme_com_b) %>%
wilcox_test(composite_mean ~ regen, detailed = TRUE) %>%
mutate(test = "Mann-Whitney U") %>%
select(theme_com_b, test, statistic, p, method, alternative, n1, n2)
view(mw_results)delta_results <- df_composites %>%
group_by(theme_com_b) %>%
summarise(
# Using effsize:: explicitly prevents conflicts
cliff = effsize::cliff.delta(composite_mean ~ regen)$estimate,
conf_low = effsize::cliff.delta(composite_mean ~ regen)$conf.int[1],
conf_high = effsize::cliff.delta(composite_mean ~ regen)$conf.int[2]
)
view(delta_results)test_table <- mw_results %>%
left_join(delta_results, by = "theme_com_b") %>%
mutate(
p_holm = p.adjust(p, method = "holm")
) %>%
arrange(theme_com_b)
test_table## # A tibble: 9 × 12
## theme_com_b test statistic p method alternative n1 n2 cliff
## <chr> <chr> <dbl> <dbl> <chr> <chr> <int> <int> <dbl>
## 1 animal_capabili… Mann… 82 0.157 Wilco… two.sided 11 11 0.355
## 2 animal_motivati… Mann… 61.5 0.974 Wilco… two.sided 11 11 0.0165
## 3 animal_opportun… Mann… 54 0.69 Wilco… two.sided 11 11 -0.107
## 4 planet_capabili… Mann… 82.5 0.152 Wilco… two.sided 11 11 0.364
## 5 planet_motivati… Mann… 61 1 Wilco… two.sided 11 11 0.00826
## 6 planet_opportun… Mann… 57.5 0.867 Wilco… two.sided 11 11 -0.0496
## 7 wellbeing_capab… Mann… 61.5 0.973 Wilco… two.sided 11 11 0.0165
## 8 wellbeing_motiv… Mann… 68.5 0.62 Wilco… two.sided 11 11 0.132
## 9 wellbeing_oppor… Mann… 69 0.594 Wilco… two.sided 11 11 0.140
## # ℹ 3 more variables: conf_low <dbl>, conf_high <dbl>, p_holm <dbl>write_csv(test_table, "COMB_OHtheme_results.csv")Tests just across COM-B components
First aggregate scores just according to COM-B, not One Health theme
df_composites_overall <- df_long %>%
group_by(id, regen, com_b) %>%
summarise(
composite_mean = mean(score, na.rm = TRUE),
composite_sum = sum(score, na.rm = TRUE),
n_items = sum(!is.na(score)),
.groups = "drop"
)Now calculate medians and IQRs for both groups across COM-B composites
# ensure consistent COM-B order in outputs/plots
df_composites_overall <- df_composites_overall %>%
mutate(
com_b = factor(com_b, levels = c("capability", "opportunity", "motivation"),
labels = c("Capability", "Opportunity", "Motivation"))
)
# Summary for composite_mean
comb_summary_mean <- df_composites_overall %>%
group_by(com_b, regen) %>%
summarise(
n = sum(!is.na(composite_mean)),
median = median(composite_mean, na.rm = TRUE),
IQR = IQR(composite_mean, na.rm = TRUE),
min = suppressWarnings(min(composite_mean, na.rm = TRUE)),
max = suppressWarnings(max(composite_mean, na.rm = TRUE)),
n_missing = sum(is.na(composite_mean)),
.groups = "drop"
) %>%
# round to two decimal places
mutate(
median = round(median, 2),
IQR = round(IQR, 2),
min = round(min, 2),
max = round(max, 2)
) %>%
arrange(com_b, regen)
# Save as CSV
write_csv(comb_summary_mean, "COMB_overall_median_IQR_mean.csv")
# print a neat table
comb_summary_mean## # A tibble: 6 × 8
## com_b regen n median IQR min max n_missing
## <fct> <fct> <int> <dbl> <dbl> <dbl> <dbl> <int>
## 1 Capability Regenerative 11 4.11 0.44 3.22 4.67 0
## 2 Capability Conventional 11 3.78 0.72 2.62 4.78 0
## 3 Opportunity Regenerative 11 3.5 0.67 3 4.25 0
## 4 Opportunity Conventional 11 3.67 0.79 2.58 4.58 0
## 5 Motivation Regenerative 11 4.04 0.26 3.26 4.3 0
## 6 Motivation Conventional 11 3.7 0.63 3.04 4.74 0mw_results_COMB <- df_composites_overall %>%
group_by(com_b) %>%
wilcox_test(composite_mean ~ regen, detailed = TRUE) %>%
mutate(test = "Mann-Whitney U") %>%
select(com_b, test, statistic, p, method, alternative, n1, n2)
view(mw_results_COMB)delta_results_COMB <- df_composites_overall %>%
group_by(com_b) %>%
summarise(
# Using effsize:: explicitly prevents conflicts
cliff = effsize::cliff.delta(composite_mean ~ regen)$estimate,
conf_low = effsize::cliff.delta(composite_mean ~ regen)$conf.int[1],
conf_high = effsize::cliff.delta(composite_mean ~ regen)$conf.int[2]
)
view(delta_results_COMB)test_table_COMB <- mw_results_COMB %>%
left_join(delta_results_COMB, by = "com_b") %>%
mutate(
p_holm = p.adjust(p, method = "holm")
) %>%
arrange(com_b)
test_table_COMB## # A tibble: 3 × 12
## com_b test statistic p method alternative n1 n2 cliff conf_low
## <fct> <chr> <dbl> <dbl> <chr> <chr> <int> <int> <dbl> <dbl>
## 1 Capabil… Mann… 79.5 0.221 Wilco… two.sided 11 11 0.314 -0.130
## 2 Opportu… Mann… 61 1 Wilco… two.sided 11 11 0.00826 -0.448
## 3 Motivat… Mann… 77 0.292 Wilco… two.sided 11 11 0.273 -0.204
## # ℹ 2 more variables: conf_high <dbl>, p_holm <dbl>write_csv(test_table_COMB, "COMB_only_results.csv")Produce violin plots for high-level COM-B only
p_comb <- ggplot(df_composites_overall,
aes(x = regen,
y = composite_mean,
fill = regen)) +
# Violin
geom_violin(trim = FALSE, alpha = 0.6, width = 0.9, colour = NA) +
# Boxplot overlay
geom_boxplot(width = 0.25, alpha = 0.7, outlier.shape = NA) +
# Individual points
geom_jitter(width = 0.05, size = 2, alpha = 0.9) +
# Facet by COM‑B component
facet_wrap(~ com_b, scales = "free_y") +
# Colours
scale_fill_brewer(palette = "Set2") +
labs(
x = "Farm type",
y = "Mean score"
) +
theme_minimal(base_size = 13) +
theme(
legend.position = "none",
strip.text = element_text(size = 12, face = "bold")
)
p_comb #prints plot
ggsave(
filename = "COMB_overall_violin.png",
plot = p_comb,
width = 10, # inches
height = 6, # inches
dpi = 300 # publication quality
)df_tidy$farming_time <- fct_relevel(
df_tidy$farming_time,
"3-5 years",
"6-10 years",
"11-15 years",
"16-20 years",
"21-30 years",
"31-40 years",
"41-50 years",
"51-60 years"
)
#create plot
p_farming_time <- ggplot(
df_tidy,
aes(
x = farming_time,
y = after_stat(count / sum(count)),
fill = farming_time
)
) +
# Bars
geom_bar(
stat = "count",
alpha = 0.7,
width = 0.9,
colour = NA
) +
# Percentage labels on bars
geom_text(
stat = "count",
aes(
label = scales::percent(
after_stat(count / sum(count)),
accuracy = 1
)
),
vjust = -0.4,
size = 3.2
) +
# Y-axis as percentages
scale_y_continuous(
labels = scales::percent_format(accuracy = 1),
expand = expansion(mult = c(0, 0.08)) # room for labels
) +
# Colours to match other figure
scale_fill_brewer(palette = "Set2") +
labs(
x = "Years of farming experience",
y = "Percentage of respondents"
) +
# Theme to match existing plot
theme_minimal(base_size = 13) +
theme(
legend.position = "none",
axis.text.x = element_text(size = 9), # smaller x-axis labels
panel.grid.major = element_blank(), # remove gridlines
panel.grid.minor = element_blank()
)
p_farming_time
ggsave(
filename = "Farming_Experience.png",
plot = p_farming_time,
width = 10, # inches
height = 6, # inches
dpi = 300 # publication quality
)