Marco Hallos

1 Data processing

pacman::p_load(
  tidyverse,
  janitor,
  summarytools,
  DataExplorer,
  readxl,
  effsize)

theme_set(theme_bw())

1.1 import data base

library(readxl)
ds <- read_excel("C:/Users/luisf/Dropbox/Puc-Rio/Consultoria - Marco Hallos/Modificado - AMBS Tabela Pedida Examinadores 31 05 2021 Oficial.xlsx")

1.2 backup

backup <- ds

1.3 clean names

Modificação dos nomes das variáveis

ds <- clean_names(ds)

1.4 ajustes das variáveis

1.5 Email do avaliador no caso a pessoa que avaliou o seu lider

criar nova variável e verificar se cada email aparece 2 ou mais vezes

ds <- ds %>% 
  mutate(email_avaliador = email)

ds %>% count(email_avaliador)

NA

Formato do email - tirar letras maúsculas

ds <- ds %>% 
  mutate(email_avaliador = tolower(email_avaliador))

1.6 Gênero do avaliador

gênero = 1 é Masculino. gênero = 2 é feminino.

ds <- ds %>% 
  mutate(sexo_avaliador= factor(if_else(genero_avaliador == 1, "male", "female")))

ds %>% count(genero_avaliador)

ds %>% count(sexo_avaliador) 

1.7 Idade do gestor

ds <- ds %>% 
  mutate(idade = idadegestor)

ds %>% count(idade)

mean(ds$idade, na.rm=T)

1.8 tempo da intervenção

valor 1 equivale ao período anterior à intervenção valor 2 e 3 equivalem ao período posterior à intervenção a diferença entre 2 e 3 é que 2 é pouco tempo após a intervenção e 3 é um tempo maior após a intervenção

Transformar em fator

ds <- ds %>% 
  mutate(tempo = if_else(versao == 1,"pre","post"))

ds <- ds %>% 
  mutate(tempo = factor(tempo, levels=c("pre","post")))

ds %>% count(tempo, versao)

1.9 Email do lider

criar nova variável e verificar se cada email aparece 2 ou mais vezes

ds <- ds %>% 
  mutate(email_lider = lider)

ds %>% 
  count(email_lider) %>% 
  arrange(desc(n))

NA

Formato do email - tirar letras maúsculas

ds <- ds %>% 
  mutate(email_lider = tolower(email_lider))

ds %>% count(email_lider) %>% 
  arrange(desc(n))

1.10 Gênero do lider

gênero = 1 é Masculino. gênero = 2 é feminino.

ds <- ds %>% 
  mutate(sexo_lider= factor(if_else(
                            genero_lider == 1,"male","female")))

ds %>% count(genero_lider)

ds %>% count(sexo_lider)

NA

1.11 Geração do Lider

Baby Boomers – 1945 – 1964 54 a 73 anos = 1 Geração X – 1965 – 1984 34 a 53 anos = 2 Geração Y – 1985 – 1999 19 a 33 anos = 3 Geração Z – 2000 – Atual Menos que 19 anos = 4

ds <- ds %>% 
  mutate(geracao = as.factor(geracao))
ds %>% count(geracao)

1.12 Senioridade do lider

ds %>% 
  distinct(email_lider, .keep_all = T) %>% #tira duplicado
  count(senioridade) %>% 
  mutate(prop = n/sum(n))

ds %>% 
    distinct(email_lider, .keep_all = T) %>% #tira duplicado
  count(geracao) %>% 
  mutate(prop = n/sum(n))

1.13 scales

1.14 Missing data within scales

ds %>% select(starts_with("v"), -versao) %>% 
  DataExplorer::plot_missing()

1.15 Compute totals

ds <- ds %>% 
  mutate(self_awareness = rowMeans(select(., v03, v07, v12, v18))) %>% 
  mutate(tranformational = rowMeans(select(., v01, v05, v09, v11, v13, v15, v17))) %>%
  mutate(emotional_intel = rowMeans(select(.,   v02, v04, v06, v08, v10, v14, v16, v19))) %>%
  mutate(lmx = rowMeans(select(., v20:v26))) 

Done!

2 Manuscript

3 Research question

3.1 Efeito da intervenção

ds %>% 
  select(email_lider, tempo,self_awareness:lmx)

ds %>% 
  select(email_lider, tempo,self_awareness:lmx) %>% 
  pivot_longer(-c(email_lider, tempo)) %>%
  mutate(name = case_when(
    name == "self_awareness" ~ "Self awareness",
    name == "tranformational" ~ "Tranformational",
    name == "emotional_intel" ~ "Emotional Intelligence",
    name == "lmx" ~ "LMX")) %>% 
  ggplot(., aes(x=tempo, y = value, group = email_lider, color = email_lider)) +
  stat_summary(geom = "line", fun = "mean") +
  stat_summary(geom = "errorbar", width = 0.1) +
  facet_wrap(~name) +
  labs(x = "", y = "") +
  theme(legend.position = "none")

3.2 Overall growth

3.3 Plot

ds %>% 
  select(email_lider, tempo,self_awareness:lmx) %>% 
  #mutate(self_awareness = scale(self_awareness)) %>% 
  #mutate(tranformational = scale(tranformational)) %>% 
  #mutate(emotional_intel = scale(emotional_intel)) %>% 
  #mutate(lmx = scale(lmx)) %>% 
  pivot_longer(-c(email_lider, tempo)) %>% 
    mutate(name = case_when(
    name == "self_awareness" ~ "Self awareness",
    name == "tranformational" ~ "Tranformational",
    name == "emotional_intel" ~ "Emotional Intelligence",
    name == "lmx" ~ "LMX")) %>% 
  ggplot(., aes(x=tempo, y = value, group = name, color = name)) +
  labs(x = "", color = "Variável", y = "Resultado") +
  stat_summary(geom = "line", fun = "mean", size = 1.5) +
  stat_summary(geom = "errorbar", width = 0.1) 

3.4 Correlation

ds %>% 
  split(.$tempo) %>% 
  map(select, self_awareness:lmx) %>% 
  map(cor)

library(igraph)
library(corrr)

cor_graph <- ds %>% 
  select(tempo,self_awareness:lmx) %>% #select all variables of intereset
  mutate(tempo = if_else(tempo == "pre",".Pre","Post")) %>% #for alignment and order
  rename("Self Awareness" = self_awareness) %>% 
  rename("Transformational" = tranformational) %>% 
  rename("Emotional Intelligence" = emotional_intel) %>% 
  rename("LMX" = lmx) %>% 
  group_by(tempo) %>% #group by age
  nest() %>% #create specific dataset to each age intervla
  mutate(data = map(data, purrr::compose(stretch, correlate))) %>% #run correlation 
  unnest(cols = -tempo) %>% 
  select(x, y, r, tempo) %>% 
  graph_from_data_frame(directed = FALSE)

library(ggraph)
ggraph(cor_graph, layout = "kk") +
  geom_edge_link(aes(edge_alpha = abs(r), color = r), edge_width = 5) +
  guides(edge_alpha = "none") +
  scale_edge_colour_gradientn(limits = c(-1, 1), colors = heat.colors(5)) +
  geom_node_point(color = "black", size = 4) +
  geom_node_text(aes(label = name), repel = TRUE) +
  facet_edges(~tempo) +
  theme_minimal() 

out_t_test <- ds %>%   
  ungroup() %>% 
  select(email_lider, tempo,self_awareness:lmx) %>% 
  pivot_longer(-c(email_lider, tempo)) %>% 
  group_by(name) %>% 
  summarise(x = list(t.test(value ~ tempo, paired=T))) 

out_t_test$x

ds %>%   
  select(email_lider, tempo,self_awareness:lmx) %>% 
  pivot_longer(-c(email_lider, tempo)) %>% 
  filter(name == "emotional_intel") %>% 
  {t.test(value ~ tempo, paired = T, data = .)}
  

3.5 paired t

 ds %>%
   select(email_lider, tempo,self_awareness:lmx) %>% 
  pivot_longer(-c(email_lider, tempo)) %>% 
  split(.$name) %>%
  map(~t.test(value ~ tempo, paired = T,.x))
#https://stackoverflow.com/questions/51074328/perform-several-t-tests-simultaneously-on-tidy-data-in-r

3.6 effect size

 ds %>%
   select(email_lider, tempo,self_awareness:lmx) %>% 
  pivot_longer(-c(email_lider, tempo)) %>% 
  split(.$name) %>%
  map(~cohen.d(value ~ tempo, data =.))

Get the summaries

ds %>%
select(email_lider, tempo,self_awareness:lmx) %>% 
  pivot_longer(-c(email_lider, tempo)) %>% 
  arsenal::tableby(interaction(name, tempo) ~ value, data = .) %>% 
  summary() 

	emotional_intel.pre (N=109)	lmx.pre (N=109)	self_awareness.pre (N=109)	tranformational.pre (N=109)	emotional_intel.post (N=109)	lmx.post (N=109)	self_awareness.post (N=109)	tranformational.post (N=109)	Total (N=872)	p value
value										< 0.001
Mean (SD)	3.810 (0.721)	3.824 (0.574)	3.624 (0.658)	4.076 (0.592)	3.972 (0.597)	4.034 (0.564)	3.867 (0.608)	4.207 (0.538)	3.927 (0.630)
Range	1.375 - 5.000	2.000 - 4.857	2.000 - 5.000	2.143 - 5.000	2.125 - 5.000	2.286 - 5.000	2.250 - 5.000	2.429 - 5.000	1.375 - 5.000

3.7 Empresas

ds <- ds %>% 
  mutate(empresa = as.factor(empresa))

ds %>% count(empresa)

3.8 Plots

ds %>% 
  filter(tempo == "pre") %>% 
  ggstatsplot::ggbetweenstats(
    data = .,
    x = empresa,
    y = lmx,
    xlab = "Empresas",
    ylab = "LMX",
    pairwise.comparisons = FALSE,
    results.subtitle = FALSE,
    title = "Resultados médios antes da intervenção"
)

ds %>% 
  filter(tempo == "pre") %>% 
  arsenal::tableby(empresa ~lmx, data = .) %>% 
  summary()

	1 (N=47)	3 (N=18)	4 (N=44)	Total (N=109)	p value
lmx					0.154
Mean (SD)	3.927 (0.474)	3.627 (0.658)	3.795 (0.623)	3.824 (0.574)
Range	2.857 - 4.857	2.000 - 4.714	2.286 - 4.857	2.000 - 4.857

ds %>% 
  select(empresa, email_lider, tempo,self_awareness:lmx) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
      mutate(name = case_when(
    name == "self_awareness" ~ "Self awareness",
    name == "tranformational" ~ "Tranformational",
    name == "emotional_intel" ~ "Emotional Intelligence",
    name == "lmx" ~ "LMX")) %>% 
  ggplot(., aes(x=tempo, y = value, group = name, color = name)) +
  stat_summary(geom = "line", fun = "mean", size = 1.5) +
  stat_summary(geom = "errorbar", width = 0.1, size = 1) +
  labs(x="", y = "Resultado", color = "Variável") +
  theme(legend.position = "bottom") +
  facet_wrap(~empresa)

ds %>% 
  select(empresa, email_lider, tempo,self_awareness:lmx) %>% 
  mutate_at(vars(self_awareness:lmx), ~scale(.)) %>% 

  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  mutate(name = case_when(
    name == "self_awareness" ~ "Self awareness",
    name == "tranformational" ~ "Tranformational",
    name == "emotional_intel" ~ "Emotional Intelligence",
    name == "lmx" ~ "LMX")) %>% 
  ggplot(., aes(x=tempo, y = value, group = name, color = name)) +
  stat_summary(geom = "line", fun = "mean", size = 1.5) +
  stat_summary(geom = "errorbar", width = 0.1, size = 1) +
  labs(x="", y = "Resultado", color = "Variável") +
  theme(legend.position = "bottom") +
  facet_wrap(~empresa)

ds %>% 
  filter(tempo == "post") %>% 
  ggstatsplot::ggbetweenstats(
    data = .,
    x = empresa,
    y = lmx,
        xlab = "Empresas",
    ylab = "LMX",
    pairwise.comparisons = FALSE,
    results.subtitle = FALSE,
    title = "Resultados médios após da intervenção"
)

3.9 paired T - Self-awareness em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo,self_awareness) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~t.test(value ~ tempo, paired = T,.x))

3.10 Média e desvio - Self-awareness em função do tempo e por empresa

ds %>%
select(empresa, email_lider, tempo,self_awareness) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  arsenal::tableby(interaction(tempo, empresa) ~ value, data = .) %>% 
  summary() 

	pre.1 (N=47)	post.1 (N=47)	pre.3 (N=18)	post.3 (N=18)	pre.4 (N=44)	post.4 (N=44)	Total (N=218)	p value
value								0.039
Mean (SD)	3.649 (0.753)	3.979 (0.667)	3.472 (0.781)	3.819 (0.427)	3.659 (0.476)	3.767 (0.596)	3.745 (0.644)
Range	2.000 - 5.000	2.250 - 5.000	2.000 - 4.500	3.250 - 4.750	2.250 - 4.250	2.250 - 4.750	2.000 - 5.000

3.11 cohen d - Self awareness em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo, self_awareness) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~cohen.d(value ~ tempo, data =.))

3.12 paired T - Tranformational em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo,tranformational) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~t.test(value ~ tempo, paired = T,.x))

3.13 Média e desvio - Tranformational em função do tempo e por empresa

ds %>%
select(empresa, email_lider, tempo, tranformational) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  arsenal::tableby(interaction(tempo, empresa) ~ value, data = .) %>% 
  summary() 

	pre.1 (N=47)	post.1 (N=47)	pre.3 (N=18)	post.3 (N=18)	pre.4 (N=44)	post.4 (N=44)	Total (N=218)	p value
value								0.228
Mean (SD)	4.143 (0.571)	4.310 (0.525)	4.000 (0.679)	4.095 (0.523)	4.036 (0.583)	4.143 (0.551)	4.142 (0.568)
Range	2.857 - 5.000	2.714 - 5.000	2.857 - 5.000	3.143 - 4.857	2.143 - 5.000	2.429 - 5.000	2.143 - 5.000

3.14 cohen d - Tranformational em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo, tranformational) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~cohen.d(value ~ tempo, data =.))

3.15 paired T - Emotional Intelligence em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo, emotional_intel) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~t.test(value ~ tempo, paired = T,.x))

3.16 Média e desvio - Emotional Intelligence em função do tempo e por empresa

ds %>%
select(empresa, email_lider, tempo, emotional_intel) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  arsenal::tableby(interaction(tempo, empresa) ~ value, data = .) %>% 
  summary() 

	pre.1 (N=47)	post.1 (N=47)	pre.3 (N=18)	post.3 (N=18)	pre.4 (N=44)	post.4 (N=44)	Total (N=218)	p value
value								0.193
Mean (SD)	3.739 (0.879)	4.008 (0.610)	3.736 (0.696)	3.757 (0.714)	3.915 (0.516)	4.023 (0.520)	3.891 (0.665)
Range	1.375 - 5.000	2.750 - 5.000	2.375 - 4.750	2.125 - 4.750	2.625 - 4.875	2.625 - 5.000	1.375 - 5.000

3.17 cohen d - Emotional Intelligence em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo, emotional_intel) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~cohen.d(value ~ tempo, data =.))

3.18 paired T - lmx em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo, lmx) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~t.test(value ~ tempo, paired = T,.x))

3.19 Média e desvio - lmx em função do tempo e por empresa

ds %>%
select(empresa, email_lider, tempo, lmx) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  arsenal::tableby(interaction(tempo, empresa) ~ value, data = .) %>% 
  summary() 

	pre.1 (N=47)	post.1 (N=47)	pre.3 (N=18)	post.3 (N=18)	pre.4 (N=44)	post.4 (N=44)	Total (N=218)	p value
value								0.006
Mean (SD)	3.927 (0.474)	4.170 (0.528)	3.627 (0.658)	3.857 (0.470)	3.795 (0.623)	3.961 (0.613)	3.929 (0.578)
Range	2.857 - 4.857	3.143 - 5.000	2.000 - 4.714	2.857 - 4.857	2.286 - 4.857	2.286 - 5.000	2.000 - 5.000

3.20 cohen d - lmx em função do tempo e por empresa

 ds %>%
   select(empresa,email_lider, tempo, lmx) %>% 
  pivot_longer(-c(empresa, email_lider, tempo)) %>% 
  split(.$empresa,) %>%
  map(~cohen.d(value ~ tempo, data =.))

3.21 Modelo hierárquico

ds %>% 
  distinct(email_lider, .keep_all = TRUE) %>% 
  count(sexo_lider) %>% 
  mutate(prop = n/sum(n)) %>% 
  adorn_totals()

library(lme4)
library(lmerTest)

mod_hier_self <- lmer(self_awareness ~ factor(tempo) + (1|email_lider), data = ds)
mod_hier_trans <- lmer(tranformational ~ factor(tempo) + (1|email_lider), data = ds)
mod_hier_intel <- lmer(emotional_intel ~ factor(tempo) + (1|email_lider), data = ds)
mod_hier_lmx <- lmer(lmx ~ factor(tempo) + (1|email_lider), data = ds)

anova(mod_hier_self, type = 3, ddf = "Satterthwaite")
anova(mod_hier_trans, type = 3, ddf = "Satterthwaite")
anova(mod_hier_intel, type = 3, ddf = "Satterthwaite")
anova(mod_hier_lmx, type = 3, ddf = "Satterthwaite")

3.22 Mediation (1)

3.23 Create a wide data

ds_wide <- ds  %>% 
  select(email_lider, email_avaliador, tempo, self_awareness:lmx) %>% 
   pivot_wider(names_from = c(tempo), values_from=c(self_awareness:lmx))  %>% 
  unchop(everything())

#jonatan teixeira avalou duas pessoas

#old code
#ds_long <- ds %>% 
#  select(email_lider, email_avaliador, tempo, self_awareness, lmx) %>% 
#   pivot_wider(names_from = c(tempo), values_from=c(self_awareness, lmx),
#                values_fn = list(. = list))  %>% 
#  unchop(everything())

library(lavaan)
model <- ' # direct effect
             lmx_post ~ c*self_awareness_pre
           # mediator
             self_awareness_post ~ a*self_awareness_pre
             lmx_post ~ b*self_awareness_post
           # indirect effect (a*b)
             ab := a*b
           # total effect
             total := c + (a*b)'

fit <- sem(model, data = ds_wide)

semPlot::semPaths(fit, "par",
             sizeMan = 15, sizeInt = 15, sizeLat = 15,
             edge.label.cex=1.5,
             fade=FALSE, rotation = 2)

summary(fit)

psych::mediate(lmx_post ~ self_awareness_pre + (self_awareness_post), data = ds_wide) 

3.24 Mediation (2)

model <- 'lmx_post ~ self_awareness_pre+lmx_pre
          self_awareness_post ~ self_awareness_pre
          lmx_post ~ self_awareness_post
          lmx_pre ~~ 0*self_awareness_pre'


fit <- sem(model, data = ds_wide)

summary(fit)

semPlot::semPaths(fit, "par",
             sizeMan = 15, sizeInt = 15, sizeLat = 15,
             edge.label.cex=1.5,
             fade=FALSE, rotation = 2)

3.25 Regressao (apos reunião)

3.26 create a wide data

ds_wide <- ds  %>% 
  select(email_lider, email_avaliador, tempo, self_awareness:lmx) %>% 
   pivot_wider(names_from = c(tempo), values_from=c(self_awareness:lmx))  %>% 
  unchop(everything())

#jonatan teixeira avalou duas pessoas

3.27 create change variables

ds_wide  <- ds_wide  %>% 
  mutate(delta_self_awareness = self_awareness_post - self_awareness_pre)

ds_wide  <- ds_wide  %>% 
  mutate(delta_tranformational = tranformational_post - tranformational_pre)

ds_wide  <- ds_wide  %>% 
  mutate(delta_emotional_intel = emotional_intel_post - emotional_intel_pre)

ds_wide  <- ds_wide  %>% 
  mutate(delta_lmx = lmx_post - lmx_pre)

ds_wide  %>% select(lmx_post,lmx_pre, delta_lmx)

3.28 linear regression (via function)

reg_aut <- function(var) {
  #model_formula <- formula(paste0(var, "~ self_awareness_pre"))
  lm(paste0(var, "~ delta_self_awareness"), data = ds_wide) %>% olsrr::ols_regress()
}

Emotional intelligence

reg_aut("delta_emotional_intel")

Transformational

reg_aut("delta_tranformational")

LMX

reg_aut("delta_lmx")

3.29 Plots

ggplot(ds_wide,aes(x = delta_self_awareness, y = delta_tranformational )) +
  geom_jitter() +
  geom_smooth(method = "lm") +
  theme_bw() + labs(x = "Self Awareness", y = "Tranformational") +
  ggpubr::stat_cor(method = "pearson")

ggplot(ds_wide,aes(x = delta_self_awareness, y = delta_emotional_intel )) +
  geom_jitter() +
  geom_smooth(method = "lm") +
  theme_bw() + labs(x = "Self Awareness", y = "Emotional intelligence") +
   ggpubr::stat_cor(method = "pearson")

ggplot(ds_wide,aes(x = delta_self_awareness, y = delta_lmx )) +
  geom_jitter() +
  geom_smooth(method = "lm") +
  theme_bw() + labs(x = "Self Awareness", y = "LMX") +
  ggpubr::stat_cor(method = "pearson")

4 Fixing inconsistencies

psych::mediate(lmx_post ~ self_awareness_pre + (self_awareness_post), data = ds_wide2) 

Done!