You are on page 1of 3

STATISTICS DAT4 EXERCISE

Introduction
In this assignment I analysed data from Earl et al. (2015). The authors’ purpose was to
prove whether circulating tumor cells (CTC) and KRAS mutant circulating free DNA
(cfDNA) detection in peripheral blood can be good markers for managing exocrine
pancreatic cancer patients.
I tried to find a good explanation for mortality considering all variables present in the
dataframe, which are:
- edad: age of the patient (quantitative).
- sexo: sex of the patient (Hombre=male, Mujer=female).
- TimeASE: time since diagnosis(quantitative).
- status: vital status of the patient (event=dead, censored=alive).
- CTC: circulating tumor cells (POS=positive, NEG=negative).
- Mutant: mutated KRAS in plasma (POS=positive, NEG=negative).
- group: clinical group (R=resectable, LA=locally advanced disease, M=stage IV
metastatic disease).

Contingency tables
Firs of all, I looked for relationships between different qualitative variables and mortality
(“status” variable), creating contingency tables with them and performing a χ2 test.
In some cases, Chi-squared tests retrieved a warning message. That happens when one
or more of the instances have an absolute frequency less than 5 and may cause the
result to be incorrect. When this happened, I performed a Fisher’s Exact test as well in
order to check the results (only for 2x2 contingency tables).
Sex vs mortality
The p-value of chi-squared test is greater than 0.05, we have no significative statistic evidence
for rejecting the null hypothesis, i.e., we cannot say that sex and mortality are dependent
variables. The Fisher’s test (although it was not needed) shows the same results as the chi-
squared test, not only the p-value, but also the Odds Ratio’s confidence interval includes number
1.

Circulating tumor cells vs mortality

The p-value of chi-squared test is greater than 0.05, we have no significative statistic evidence
for rejecting the null hypothesis, i.e., we cannot say that circulating tumor cells and mortality
are dependent variables. Moreover, the Fisher’s test is needed, and it shows the same results
as the chi-squared test, not only the p-value, but also the Odds Ratio’s confidence interval
includes number 1.

Clinical group vs mortality


The p-value of chi-squared test is less than 0.05, we have significative statistic evidence for
rejecting the null hypothesis, i.e., we can say that the clinical group and mortality are related
variables in some manner. The Fishers’ exact test cannot be applied to this contingency table
even though chi-squared’s approximation may be incorrect.

Mutant KRAS cfDNA vs mortality

The p-value of chi-squared test is greater than 0.05, we have no significative statistic evidence
for rejecting the null hypothesis, i.e., we cannot say that having mutant KRAS cfDNA and
mortality are dependent variables. Moreover, the Fisher’s test is needed, and it shows the same
results as the chi-squared test, not only the p-value, but also the Odds Ratio’s confidence interval
includes number 1.

Logistic Regression

You might also like