MSc_PH_Project

Load packages

The following R packages were used for this analysis:

tidymodels
tidyverse
plotly
pacman
skimr
ggplot2
forcats
themis
readxl
janitor
klaR
tune
discrim
vip
AMR
party

msc <- read_csv("mscph.csv", col_types = cols(),  na = "NA")
dim(msc)

[1] 600  17

msc %>% count(final_outcome_result)

# A tibble: 2 × 2
  final_outcome_result     n
  <chr>                <int>
1 HIV Infected           210
2 HIV Uninfected         390

new_msc <- msc %>% select(-first_pcr_test_result)
skimr::skim(new_msc)

Data summary
Name	new_msc
Number of rows	600
Number of columns	16
_______________________
Column type frequency:
character	12
numeric	4
________________________
Group variables	None

Variable type: character

skim_variable	n_missing	complete_rate	min	max	n_unique
pep_id_sex	0	1.00	4	6	2
pep_id_deliv_type	0	1.00	16	18	4
pep_id_hiv_serostatus_mo	0	1.00	14	22	3
second_pcr_test	0	1.00	8	8	2
final_outcome_result	0	1.00	12	14	2
mother_received_intervetion_y_n	0	1.00	2	3	2
place_site_of_delivery	0	1.00	15	16	2
brst_feeding_method	0	1.00	9	13	2
type_of_avr_prophy_for_baby	0	1.00	3	9	2
cotrimoxazole_administered_y_n	0	1.00	2	3	2
education	0	1.00	7	9	3
marital_st	85	0.86	6	9	5

Variable type: numeric

skim_variable	complete_rate	mean	sd	p0	p25	p50	p75	p100	hist
pep_id_gest_age_birth_wks	1	38.05	0.51	36	38	38	38	40	▁▁▇▁▁
pep_id_birth_weight_grams	1	3043.15	581.61	2000	2600	3000	3400	5700	▆▇▃▁▁
mothers_recent_viral_load	1	3620.24	21581.18	0	20	20	45	234542	▇▁▁▁▁
age_mo	1	37.38	6.06	22	33	38	41	69	▂▇▃▁▁

Exploratory analysis

# Change variables from character to factor
new_msc$pep_id_sex <- factor(new_msc$pep_id_sex)

new_msc$pep_id_deliv_type <- factor(new_msc$pep_id_deliv_type)

new_msc$second_pcr_test <- factor(new_msc$second_pcr_test)

new_msc$final_outcome_result <- factor(new_msc$final_outcome_result)

new_msc$mother_received_intervetion_y_n <- factor(new_msc$mother_received_intervetion_y_n)

new_msc$brst_feeding_method <- factor(new_msc$brst_feeding_method)

new_msc$type_of_avr_prophy_for_baby <- factor(new_msc$type_of_avr_prophy_for_baby)

new_msc$cotrimoxazole_administered_y_n <- factor(new_msc$cotrimoxazole_administered_y_n)

new_msc$pep_id_hiv_serostatus_mo <- factor(new_msc$pep_id_hiv_serostatus_mo)

new_msc$place_site_of_delivery <- factor(new_msc$place_site_of_delivery)

new_msc$education <- factor(new_msc$education)

new_msc$marital_st <- factor(new_msc$marital_st)

# Variable sex
new_msc %>% count(pep_id_sex)

## # A tibble: 2 × 2
##   pep_id_sex     n
##   <fct>      <int>
## 1 Female       289
## 2 Male         311

# HIV status of mother
new_msc %>% count(pep_id_hiv_serostatus_mo)

## # A tibble: 3 × 2
##   pep_id_hiv_serostatus_mo     n
##   <fct>                    <int>
## 1 HIV-1 & HIV-2 positive      69
## 2 HIV-1 positive             521
## 3 HIV-2 positive              10

# Final outcome of infants
new_msc %>% count(final_outcome_result)

## # A tibble: 2 × 2
##   final_outcome_result     n
##   <fct>                <int>
## 1 HIV Infected           210
## 2 HIV Uninfected         390

# Mode of delivery
new_msc %>%  count(pep_id_deliv_type)

## # A tibble: 4 × 2
##   pep_id_deliv_type      n
##   <fct>              <int>
## 1 Assisted vaginal      12
## 2 Elective cesarean    137
## 3 Emergency cesarean    44
## 4 Standard vaginal     407

# PCR count of infants
new_msc %>% count(second_pcr_test)

## # A tibble: 2 × 2
##   second_pcr_test     n
##   <fct>           <int>
## 1 Negative          577
## 2 Positive           23

# Mothers treatment
new_msc %>% count(mother_received_intervetion_y_n)

## # A tibble: 2 × 2
##   mother_received_intervetion_y_n     n
##   <fct>                           <int>
## 1 No                                 34
## 2 Yes                               566

# Delivery site
new_msc %>% count(place_site_of_delivery)

## # A tibble: 2 × 2
##   place_site_of_delivery     n
##   <fct>                  <int>
## 1 Inside Facility          571
## 2 Outside Facility          29

# Mode of feeding
new_msc %>% count(brst_feeding_method)

## # A tibble: 2 × 2
##   brst_feeding_method     n
##   <fct>               <int>
## 1 Exclusive             332
## 2 Mixed Feeding         268

# AVT for infants 
new_msc %>% count(type_of_avr_prophy_for_baby)

## # A tibble: 2 × 2
##   type_of_avr_prophy_for_baby     n
##   <fct>                       <int>
## 1 AZT + NVP                      54
## 2 NVP                           546

# Antibiotic therapy for mothers
new_msc %>% count(cotrimoxazole_administered_y_n)

## # A tibble: 2 × 2
##   cotrimoxazole_administered_y_n     n
##   <fct>                          <int>
## 1 No                                44
## 2 Yes                              556

# Mothers education level
new_msc %>% count(education)

## # A tibble: 3 × 2
##   education     n
##   <fct>     <int>
## 1 Primary      66
## 2 Secondary   367
## 3 Tertiary    167

# Marital status
new_msc %>% count(marital_st)

## # A tibble: 6 × 2
##   marital_st     n
##   <fct>      <int>
## 1 Divorced      14
## 2 Married      285
## 3 Separated     21
## 4 Single       191
## 5 Widowed        4
## 6 <NA>          85

# Infant sex
new_msc %>% count(pep_id_sex)

## # A tibble: 2 × 2
##   pep_id_sex     n
##   <fct>      <int>
## 1 Female       289
## 2 Male         311

# Age group of mothers
mothers <- msc$age_mo
group_mothers <- age_groups(mothers, split_at = "tens")
table(group_mothers)

## group_mothers
##   0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90-99  100+ 
##     0     0    50   323   217     4     6     0     0     0     0

Graphics

# Viral load of mothers
msc %>% 
  ggplot(aes(mothers_recent_viral_load)) + geom_histogram(fill = "red") + labs(x = "Viral load of mothers (copies/ml)", caption = "Histogram of mothers viral load. Most of the mothers have very low viral load but some have very high viral load")

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

# Gestational age of infants
msc %>% 
  ggplot(aes(pep_id_gest_age_birth_wks)) + geom_histogram(fill = "purple") + labs(x = "Gestational age of infants (weeks)", caption = "Histogram of infants gestational age at birth. The average age was 38 weeks")

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

# Weight of infants
msc %>% 
  ggplot(aes(pep_id_birth_weight_grams)) + geom_histogram(fill = "violet") + labs(x = "Weight of infants (grams)", caption = "Histogram showing the distribution of infants weight at birth. The majority of infants weighed between 2 to 4.5kg")

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

# Age of mothers
msc %>% 
  ggplot(aes(age_mo)) + geom_histogram(fill = "yellow") + labs(x = "Age of mothers (years)", caption = "Distribution of mothers age. The histogram is multimodal")

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

# Viral load boxplot
vload_spread <- ggplot(msc) + aes(x = "", y = mothers_recent_viral_load, na.rm = TRUE) + geom_boxplot(fill = "maroon") + labs(y = "Viral load (copies/ml)", caption = "Box plot of mothers viral load showing the spiking up of some of the viral load")
ggplotly(vload_spread)

# Box plot for infant weight
infant_wt <- ggplot(msc) + aes(x = "", y = pep_id_birth_weight_grams, na.rm = TRUE) + geom_boxplot(fill = "navy") + labs(y = "Birth weight (grams)", caption = "Box plot of infants weight. Maximum weight was 5.7kg and minimum 2kg")
ggplotly(infant_wt)

# Gestational age @ birth
age_wk <- ggplot(msc) + aes(x = "", y = pep_id_gest_age_birth_wks, na.rm = TRUE) + geom_boxplot(fill = "limegreen") + labs(y = "Gestational age (weeks)", caption = "Box plot of gestational age of infants at birth. Minimum 36 weeks, maximum 40 weeks")
ggplotly (age_wk)

# Box plot of infant weight by outcome
infant_out <- ggplot(msc) + aes(x = "", y = pep_id_birth_weight_grams, fill = final_outcome_result, na.rm = TRUE) + geom_boxplot() + labs(y = "Birth weight per outcome", caption = "Box plot showing weight distribution of infants with final outcome of HIV infected ")
ggplotly(infant_out)

# Box plot of infant weight by outcome
infant_out <- ggplot(msc) + aes(x = "", y = pep_id_birth_weight_grams, fill = final_outcome_result, na.rm = TRUE) + geom_boxplot() + labs(y = "Birth weight per outcome", caption = "Box plot showing weight distribution of infants with final outcome of HIV uninfected ")
ggplotly(infant_out)

# ARV drug by final outcome
table(msc$type_of_avr_prophy_for_baby, msc$final_outcome_result)

           
            HIV Infected HIV Uninfected
  AZT + NVP           42             12
  NVP                168            378

# Delivery mode by final outcome
table(msc$pep_id_deliv_type, msc$final_outcome_result)

                    
                     HIV Infected HIV Uninfected
  Assisted vaginal              6              6
  Elective cesarean            39             98
  Emergency cesarean           12             32
  Standard vaginal            153            254

Modeling

# Remove columns with NAs
new_msc <- new_msc %>% select(-marital_st)

# Split data
set.seed(111)
data_split <- initial_split(new_msc, prop = .8, strata = final_outcome_result)
data_split

<Training/Testing/Total>
<480/120/600>

# Training and test data
m_train <- data_split %>% training()
m_test <- data_split %>% testing()

# Recipe
reci_m <- recipe(final_outcome_result ~ ., data = m_train) %>% 
  step_rose(final_outcome_result) 
  

# Prep and juice for test
reci_prep <- reci_m %>% prep()
reci_juice <- reci_prep %>% juice()

Random forest

# Random forest
rf <- rand_forest(
  mtry = 4,
  trees = 1000,
  min_n = 1
) %>% 
  set_mode("classification") %>% 
  set_engine("ranger")

# Workflow
rf_wf <- workflow() %>% 
  add_recipe(reci_m) %>% 
  add_model(rf)

# Cross validation
set.seed(222)
m_folds <- vfold_cv(m_train)

# Set metrics
ev_metrics <- metric_set(accuracy, sensitivity, specificity, precision, roc_auc, npv)

# Evaluate model
doParallel::registerDoParallel()
set.seed(333)
rf_rs <- fit_resamples(
  rf_wf,
  resamples = m_folds,
  metrics = ev_metrics
)

collect_metrics(rf_rs)

# A tibble: 6 × 6
  .metric     .estimator  mean     n std_err .config             
  <chr>       <chr>      <dbl> <int>   <dbl> <chr>               
1 accuracy    binary     0.733    10  0.0191 Preprocessor1_Model1
2 npv         binary     0.743    10  0.0290 Preprocessor1_Model1
3 precision   binary     0.729    10  0.0523 Preprocessor1_Model1
4 roc_auc     binary     0.767    10  0.0230 Preprocessor1_Model1
5 sensitivity binary     0.407    10  0.0619 Preprocessor1_Model1
6 specificity binary     0.915    10  0.0219 Preprocessor1_Model1

Logistic regression

lr <- logistic_reg() %>% 
  set_engine(engine = "glm") %>% 
  set_mode("classification")

# Workflow
lr_wf <- workflow() %>%
  add_recipe(reci_m) %>% 
  add_model(lr)

# Evaluate
set.seed(444)
lr_rs <- fit_resamples(
  lr_wf,
  resamples = m_folds,
  metrics = ev_metrics
)

collect_metrics(lr_rs)

# A tibble: 6 × 6
  .metric     .estimator  mean     n std_err .config             
  <chr>       <chr>      <dbl> <int>   <dbl> <chr>               
1 accuracy    binary     0.740    10  0.0174 Preprocessor1_Model1
2 npv         binary     0.837    10  0.0282 Preprocessor1_Model1
3 precision   binary     0.602    10  0.0190 Preprocessor1_Model1
4 roc_auc     binary     0.778    10  0.0158 Preprocessor1_Model1
5 sensitivity binary     0.754    10  0.0263 Preprocessor1_Model1
6 specificity binary     0.732    10  0.0206 Preprocessor1_Model1

Logistic_m <- glm(
  final_outcome_result ~ ., 
  data = m_train,
  family = "binomial"
)
summary(Logistic_m)


Call:
glm(formula = final_outcome_result ~ ., family = "binomial", 
    data = m_train)

Coefficients:
                                         Estimate Std. Error z value Pr(>|z|)
(Intercept)                             1.229e+01  9.210e+00   1.335  0.18192
pep_id_sexMale                          2.365e-01  2.404e-01   0.984  0.32510
pep_id_deliv_typeElective cesarean      7.328e-01  1.008e+00   0.727  0.46710
pep_id_deliv_typeEmergency cesarean     1.165e+00  1.077e+00   1.082  0.27941
pep_id_deliv_typeStandard vaginal       2.764e-01  9.851e-01   0.281  0.77904
pep_id_gest_age_birth_wks              -1.761e-01  2.356e-01  -0.748  0.45469
pep_id_hiv_serostatus_moHIV-1 positive -3.014e+00  1.036e+00  -2.909  0.00363
pep_id_hiv_serostatus_moHIV-2 positive  1.365e+01  1.338e+03   0.010  0.99186
pep_id_birth_weight_grams              -1.162e-03  2.306e-04  -5.040 4.65e-07
mothers_recent_viral_load              -8.160e-06  9.375e-06  -0.870  0.38408
second_pcr_testPositive                -1.811e+01  7.571e+02  -0.024  0.98091
mother_received_intervetion_y_nYes     -1.782e+00  1.337e+00  -1.332  0.18285
place_site_of_deliveryOutside Facility -2.486e-01  1.092e+00  -0.228  0.82001
brst_feeding_methodMixed Feeding        1.056e+00  2.556e-01   4.131 3.61e-05
type_of_avr_prophy_for_babyNVP          2.620e+00  1.259e+00   2.081  0.03743
cotrimoxazole_administered_y_nYes       5.235e-02  4.537e-01   0.115  0.90814
age_mo                                  1.533e-03  1.984e-02   0.077  0.93841
educationSecondary                     -2.512e-01  3.928e-01  -0.640  0.52244
educationTertiary                      -1.376e-01  4.223e-01  -0.326  0.74450
                                          
(Intercept)                               
pep_id_sexMale                            
pep_id_deliv_typeElective cesarean        
pep_id_deliv_typeEmergency cesarean       
pep_id_deliv_typeStandard vaginal         
pep_id_gest_age_birth_wks                 
pep_id_hiv_serostatus_moHIV-1 positive ** 
pep_id_hiv_serostatus_moHIV-2 positive    
pep_id_birth_weight_grams              ***
mothers_recent_viral_load                 
second_pcr_testPositive                   
mother_received_intervetion_y_nYes        
place_site_of_deliveryOutside Facility    
brst_feeding_methodMixed Feeding       ***
type_of_avr_prophy_for_babyNVP         *  
cotrimoxazole_administered_y_nYes         
age_mo                                    
educationSecondary                        
educationTertiary                         
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 621.55  on 479  degrees of freedom
Residual deviance: 459.08  on 461  degrees of freedom
AIC: 497.08

Number of Fisher Scoring iterations: 16

exp(cbind(Odds_ratio = coef(Logistic_m), confint(Logistic_m)))

Waiting for profiling to be done...

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: algorithm did not converge

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: algorithm did not converge

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

Warning in regularize.values(x, y, ties, missing(ties), na.rm = na.rm):
collapsing to unique 'x' values

                                         Odds_ratio         2.5 %        97.5 %
(Intercept)                            2.184062e+05  3.101120e-03  1.794986e+13
pep_id_sexMale                         1.266852e+00  7.920845e-01  2.035661e+00
pep_id_deliv_typeElective cesarean     2.080926e+00  2.217325e-01  1.388428e+01
pep_id_deliv_typeEmergency cesarean    3.206970e+00  3.115883e-01  2.513343e+01
pep_id_deliv_typeStandard vaginal      1.318363e+00  1.454594e-01  8.402283e+00
pep_id_gest_age_birth_wks              8.385166e-01  5.269806e-01  1.333996e+00
pep_id_hiv_serostatus_moHIV-1 positive 4.907288e-02  2.701280e-03  2.421520e-01
pep_id_hiv_serostatus_moHIV-2 positive 8.437007e+05  7.306479e+14 2.889471e+235
pep_id_birth_weight_grams              9.988382e-01  9.983719e-01  9.992768e-01
mothers_recent_viral_load              9.999918e-01  9.999625e-01  1.000006e+00
second_pcr_testPositive                1.359223e-08 6.301266e-140  1.141085e-42
mother_received_intervetion_y_nYes     1.683693e-01  6.657543e-03  1.881591e+00
place_site_of_deliveryOutside Facility 7.799266e-01  8.026426e-02  6.668719e+00
brst_feeding_methodMixed Feeding       2.874948e+00  1.758467e+00  4.800733e+00
type_of_avr_prophy_for_babyNVP         1.374059e+01  1.525412e+00  3.265184e+02
cotrimoxazole_administered_y_nYes      1.053746e+00  4.239856e-01  2.543926e+00
age_mo                                 1.001534e+00  9.632457e-01  1.041533e+00
educationSecondary                     7.778291e-01  3.494402e-01  1.647425e+00
educationTertiary                      8.714253e-01  3.717511e-01  1.965514e+00

Decision trees

dt <- decision_tree() %>% 
  set_engine(engine = "rpart") %>% 
  set_mode("classification")

# Workflow
dt_wf <- workflow() %>%
  add_recipe(reci_m) %>% 
  add_model(dt)

# Evaluate
set.seed(555)
dt_rs <- fit_resamples(
  dt_wf,
  resamples = m_folds,
  metrics = ev_metrics
)

collect_metrics(dt_rs)

# A tibble: 6 × 6
  .metric     .estimator  mean     n std_err .config             
  <chr>       <chr>      <dbl> <int>   <dbl> <chr>               
1 accuracy    binary     0.7      10  0.0255 Preprocessor1_Model1
2 npv         binary     0.709    10  0.0327 Preprocessor1_Model1
3 precision   binary     0.722    10  0.0635 Preprocessor1_Model1
4 roc_auc     binary     0.637    10  0.0324 Preprocessor1_Model1
5 sensitivity binary     0.278    10  0.0659 Preprocessor1_Model1
6 specificity binary     0.940    10  0.0201 Preprocessor1_Model1

Naive Bayes

nb <- naive_Bayes() %>% 
  set_engine(engine = "klaR") %>% 
  set_mode("classification")

# Workflow
nb_wf <- workflow() %>%
  add_recipe(reci_m) %>% 
  add_model(nb)

# Evaluate
set.seed(777)
nb_rs <- fit_resamples(
  nb_wf,
  resamples = m_folds,
  metrics = ev_metrics
)

collect_metrics(nb_rs)

# A tibble: 6 × 6
  .metric     .estimator  mean     n std_err .config             
  <chr>       <chr>      <dbl> <int>   <dbl> <chr>               
1 accuracy    binary     0.721    10 0.0206  Preprocessor1_Model1
2 npv         binary     0.708    10 0.0244  Preprocessor1_Model1
3 precision   binary     0.822    10 0.0478  Preprocessor1_Model1
4 roc_auc     binary     0.804    10 0.0151  Preprocessor1_Model1
5 sensitivity binary     0.259    10 0.0197  Preprocessor1_Model1
6 specificity binary     0.972    10 0.00712 Preprocessor1_Model1

# Trees
x <- ctree(final_outcome_result ~ ., data=new_msc)
plot(x, type="extended")

cforest(final_outcome_result ~ ., data=new_msc, controls=cforest_control(mtry=4, mincriterion=0))


     Random Forest using Conditional Inference Trees

Number of trees:  500 

Response:  final_outcome_result 
Inputs:  pep_id_sex, pep_id_deliv_type, pep_id_gest_age_birth_wks, pep_id_hiv_serostatus_mo, pep_id_birth_weight_grams, mothers_recent_viral_load, second_pcr_test, mother_received_intervetion_y_n, place_site_of_delivery, brst_feeding_method, type_of_avr_prophy_for_baby, cotrimoxazole_administered_y_n, age_mo, education 
Number of observations:  600

MSc_PH_Project

Olusola Afuwape

2023-07-13

Load packages

Exploratory analysis

Graphics

Modeling

Random forest

Logistic regression

Decision trees

Naive Bayes