Report 3

Introduction

The same subsets and variables explored on previous reports were explored, but this time the matching was done over the BMI class, age and sex of the participants. Since there weren’t enough different cases to match all of them, the sample size of the original and matched sets is sometimes different, as shown on the corresponding tables.

Data Analysis

First we load our library and the file containing the results of the matching.

source('statis.R')
load('Experiment2/StaTab.RData')

Then we procceed to analyse each variable group across the different subsets.

Demographics

Comparing Descriptive Statistics

Demo = StaTab[,"Demo"]
list(
  Data = lapply(Demo, . %>% .@DataDS) %>% bind_rows(.id = "Subset"),
  Match = lapply(Demo, . %>% .@MatchDS) %>% bind_rows(.id = "Subset")  
) %>% bind_rows(.id = "Origin") %>% arrange(Var, Subset)

Wilcoxon Test Results

If we look at the results for the Demographics group, we find that BMI and weight [1, 5] are not consistently different (which is expected as we matched on the BMI classes), but mental and physical health [3, 4] are. We can also observe that drinking [2] doesn’t have an influence in most cases.

lapply(Demo, . %>% .@MatchTest) %>% bind_rows(.id = "Subset") %>% arrange(Var)

NUT

Comparing Descriptive Statistics

NUT = StaTab[,"NUT"]
list(
  Data = lapply(NUT, . %>% .@DataDS) %>% bind_rows(.id = "Subset"),
  Match = lapply(NUT, . %>% .@MatchDS) %>% bind_rows(.id = "Subset")  
) %>% bind_rows(.id = "Origin") %>% arrange(Var, Subset)

Wilcoxon Test Results

Most NUT variables don’t have a significant difference, with minor exceptions. Cheese and fruit [9, 18] consumption however, present differences in half of the cases.

lapply(NUT, . %>% .@MatchTest) %>% bind_rows(.id = "Subset") %>% arrange(Var)

NUR

Comparing Descriptive Statistics

NUR = StaTab[,"NUR"]
list(
  Data = lapply(NUR, . %>% .@DataDS) %>% bind_rows(.id = "Subset"),
  Match = lapply(NUR, . %>% .@MatchDS) %>% bind_rows(.id = "Subset")  
) %>% bind_rows(.id = "Origin") %>% arrange(Var, Subset)

Wilcoxon Test Results

On NUR variables we don’t observe a consistent behaviour, we observe equal ammounts of significant and unsignificant variables along the different subsets. However, if we take into consideration the way each of them is weighted by looking at the NUR score and the corresponding High Nutritional Risk Diagnosis [5, 3] we observe a significant difference for most subsets. This is consistent with the difference in the mean score of each subset we observe in the previous table.

lapply(NUR, . %>% .@MatchTest) %>% bind_rows(.id = "Subset") %>% arrange(Var)