Detecting Potential Outliers PDF

Preview

Summary

Lecture Notes Week 5...

Description

Detecting Potential Outliers -

multiply by 1.5 (no reason for this number)

**The interquartile range (IQR) = Q3 – Q1 IQR = Q3–Q1 

It is possible for an observation to fall outside of these bounds and not truly be an outlier (i.e., not separated by a long gap from the rest of the data).

Histogram vs. Boxplot 

Same graph in histogram and box plot - (you’re only dealing with 1 variable with both)



All observations are distributed approx equally



Q1 - left side



Q2 – median



Q3 - right side



determine max and min by checking IQR (min and max will not be out of range)

** know how to match histogram and box plots -

Left skewed histogram

-

Boxplot reveals outliers

-

You see there are 6 outliers on left side

-

Also, right side of box is smaller than left – most of our observations are on this side

-

Right skewed histogram

Associations The Data

These 6 variable – categorical, quantitative (+ continuous), quantitative (+ continuous), quantitative (+ continuous), categorical, categorical Response vs. Explanatory Variable In data analysis, we are generally interested in how the outcome or the response variable depends on or is explained by an explanatory variable. 

Response and explanatory variables can be either categorical or quantitative



There is an association between the response and an explanatory variable when there is a relationship between the two variables



If there is no association present, we say that the response and explanatory variables are independent.

Two quantitative variables: Is there an association? Scatterplot – use when you have 2 quantitative variables Which has an association? Left graph - there is association between weight and BMI, there isn’t an association between height and BMI (that’s why data is all over the place) One quantitative and one categorical variable: Is there an association? Is there an association? 

Sex doesn’t help determine BMI, sex can help you understand the variable height

**Seven Rules for Making Good Figures 1. Check the data

2. Explain color/size/symbols 3. Label axes 4. Include units 5. Keep geometry in check 6. Include your sources 7. Consider your audience Types of Study Designs -

In an experimental study researchers assign subjects to experimental conditions and then the response variable or outcome of interest is observed. The experimental conditions can be called treatments.

-

In an observational study researchers observe both the response and explanatory variable without assigning a ‘treatment’. Observational studies are non-experimental.

-

We can study the effect of an explanatory variable on a response variable more accurately in an experimental study than an observational study.

observational — measure based on prior ability experimental — random –more about cause and effect...