Last bits of inferential statistics
Dec 13, 2015
Last bits of inferential statistics
Check in
• Proposal DRAFT due Tuesday
• Research in media assignment Tuesday
• Quiz Tuesday the 29th
– Will cover statistics material– Multiple choice
Conducting statistical tests
• Testing our hypotheses numerically
• Do the data support the null distribution?
• Do the data support the alternate distribution?
• Tests tell us if the difference in distributions is significantly different– Not just look different, but different enough so
we are pretty sure its not an accident
Some basic tests
• Remember there are many different types of distributions
• The normal distribution and others
• Tests are often named for their distributions
Tests—first look
• Chi-Squared– The X2 distribution
• T-test– The students T distribution
• Rho– The correlation coefficient R– Assumes a normal distribution or T
distribution
How to choose which test?
• Many tests may answer the same type of question
• Which test you use depends on your questions and variables
• Are your variables continuous or categorical?
• May also matter if independent variable is one type, and dependent variable is another type
What type of variable do I have
• Independent variable– The predictor– E.g. If I believe that gender predicts income,
then gender is the IV
• Dependent variable– The outcome– E.g. if I believe that anxiety predicts eating
habits, then eating habits is the DV
What type of variable do I have?
• Not all questions have an independent and dependent variable
• If I’m asking if two things are ASSOCIATED or CORRELATED there is no actual IV or DV
• A variable may be an IV in one question and a DV in another question, and vice versa
• It depends on my question
Continuous vs. categorical?
• Continuous– May actually be based on ordinal– When I have a large range of numerical
ratings– Height, weight– Symptoms, feelings
Continuous vs. categorical
• Categorical– May have levels– Has few categories– 1,2,3– Yes/no– Red/blue
Which test?
• If my question is– Are these two groups different?
• And my outcome is – Categorical
• I use a X2 test
Differences between 2 groups
• IV is 2 groups– E.g. men and women– Catholics and Protestants– People who have a peanut allergy and people
who don’t
• So I know that the IV is categorical
Differences between 2 groups
• If DV is also categorical
• Will also be a sort of “grouped” variable– Employed/unemployed– Married/not married– College degree/no college degree
Chi Squared Test
• If our predictor is categorical
• And our outcome is categorical
• Then we use a chi-squared test– A “2 x 2 table approach”
Example
• Does gender predict employment– Gender is categorical– Employment is categorical– This is a chi-squared test
• Tells us if there is a difference in the expected rate of occurrence, and the observed rate of occurrence
Example
• Assume gender is 50/50• If gender does NOT predict employment,
then we expect equal numbers of men and women to be employed
• If gender DOES predict employment we observe unequal numbers of men and women employed
• Chi square tests if observed and expected are significantly different
Which test?
• If my question is are these 2 groups different
• And my outcome is continuous
• I use a T-test
T-tests
• Also a test of difference between groups
• When the IV is categorical
• And the DV is continuous
• E.g. IV—men/women
• DV—height in inches
T-tests-examples
• IV—Dogs/Cats
• DV—weight in pounds and ounces
• IV—employed/unemployed
• DV—income in dollars
• IV—full time student/not full time
• DV—average hours of sleep per night
Which test?
• If my question is are these characteristics related or unrelated
• I am looking at the relation between two continuous variables
• I use a correlation test
Correlation
• Not a test of differences between groups
• A test of whether two continuous variables are related or unrelated
• The statistic is R
Examples
• Height and Weight
• Grade point average and hours of study
• Hours of sleep and scale measure of exhaustion