Ciencia

STATISTICS: AN INTRODUCTION USING R By M.J. Crawley Exercises

3. STATISTICS OF ONE AND TWO SAMPLES

The hardest part of any statistical work is knowing how to get started. One of the hardestthings about getting started in choosing the right kind of statistical analysis for your data and your particular scientific question. The truth is that there is no substitute for experience. The way toknow what to do is to have done it properly lots of times before. But here are some useful guidelines. It is essential, for example to know: 1) Which of your variables is the response variable? 2)Which are the explanatory variables? 3) Are the explanatory variables continuous or categorical, or a mixture of both? 4) What kind of response variable have you got: is it a continuous measurement, acount, a proportion, a time-at-death or a category ? The answers to these questions should lead you quickly to the appropriate choice if you use the following dichotomous key. It begins by askingquestions about the nature of your explanatory variables, and ends by asking about what kind of response variable you have got. 1. Explanatory variables all categorical At least one explanatory variable acontinuous measurement 2. Response variable a count Response variable not a count 3. Response variable a continuous measurement Response variable other than this Contingency table 3 Analysis ofvariance Analysis of deviance 2 4

4. All explanatory variables continuous Regression 5 Explanatory variables both continuous and categorical Analysis of covariance 5 5. Response variable continuousResponse variable a count Response variable a proportion Response variable a time-at-death Response variable binary Explanatory variable is time Regression or Ancova Log-linear models (Poisson errors)Logistic model (binomial errors) Survival analysis Binary logistic analysis Time series analysis

Estimating parameters from data Data have a number of important properties, and it is useful to be...