Statistical Screening

Lendo os dados

Você pode baixar os dados aqui.

sales <- read_xlsx("dados/Stella_export_test.xlsx")

Tratando os dados

sales_long <- sales %>% 
  pivot_longer(2:dim(sales)[2], names_to = "parameter", values_to = "value") %>% 
  separate_wider_delim(parameter, delim = ":",
                       names = c("run","parameter"))%>%
  mutate(parameter = as.factor(parameter))


sales_wider <- sales_long %>% 
  pivot_wider(names_from = "parameter") 

sales_force_rnd <- runif(dim(sales_wider)[1], min = 0.001, max = 0.01)

Após a limpeza dos dados, a tabela final ficará assim:

sales_wider %>% 
  head(10) %>% 
  knitr::kable() 

Years	run	exit rate	average salary	maximum effectiveness	saturation size	price	initial sales force	Size of sales force
0	Run 1	0.247	26500	1.59	750	99.88	52.5	52.5
0	Run 2	0.215	20300	2.29	590	98.84	33.5	33.5
0	Run 3	0.239	24700	1.87	1430	101.72	35.5	35.5
0	Run 4	0.217	22100	1.91	1330	101.56	45.5	45.5
0	Run 5	0.179	25500	2.09	1030	100.84	53.5	53.5
0	Run 6	0.235	29700	1.51	850	99.40	48.5	48.5
0	Run 7	0.175	25300	1.95	650	100.52	40.5	40.5
0	Run 8	0.221	25100	2.13	1450	101.64	27.5	27.5
0	Run 9	0.205	23700	2.17	670	100.68	55.5	55.5
0	Run 10	0.225	21100	2.07	830	99.32	54.5	54.5

Calculando os coeficientes de correlação

sales_wider2 <- sales_wider %>% 
  mutate(year_factor = as.factor(Years)) %>% 
  mutate(sales_force = sales_force_rnd+` Size of sales force`) %>% 
  group_by(Years) %>%
  summarize(exit_rate = cor(` exit rate`,
                            sales_force),
            ave_salary = cor(` average salary`, sales_force),
            max_effec = cor(` maximum effectiveness`, sales_force),
            sat_size = cor(` saturation size`, sales_force),
            price = cor(` price`, sales_force),
            initial_sales = cor(` initial sales force`, sales_force))

Criando o gráfico

sales_wider2 %>% 
  pivot_longer(2:7) %>% 
  ggplot(aes(Years, value, color=name))+
  geom_line(linewidth = 1)+
  #geom_smooth()+
  scale_y_continuous(limits = c(-1,1),
                     n.breaks = 10)+
  scale_x_continuous(limits = c(0,20),
                     n.breaks = 21)+
  theme_bw()