SOCECOL 13 Chapter Notes - Chapter 10: Extrapolation, Time Series, Statistical Significance
7 May 2018
School
Department
Course
Professor
Jessica Mangold
SE13 Statistical Analysis
Professor Wong
5/6/18
Chapter 10: Relationships Between Measurement Variables
10.1 Statistical Relationships
- Introduction
- how relationships between variables can be expressed quantitatively
- correlation = measures the strength of a certain type of relationship between
two measurement variables
- regression = numerical method for trying to predict the value of one
measurement variable from knowing the value of another one
- Statistical Relationships versus Deterministic Relationships
- statistical relationship = natural variability exists in the relationship between
the two measurements
- deterministic relationship = if we know the value of one, we can determine the
value of the other exactly
- Natural Variability in Statistical Relationships
- natural variability exists in relationship between two measurements in a
statistical relationship
- statistical relationships useful for describing what happens to a pop. or aggregate
- stronger relationship -> more useful for predicting what will happen for
an individual
10.2 Strength versus Statistical Significance
- Introduction
- to find statistical relationship -> researchers usually rely on measurements from
sample of individuals
- Defining Statistical Significance
- statistically significant = chances that a relationship that’s strong or stronger
would have been observed in sample if there really were nothing going on in the
population
- if chances are small -> then relationship is statistically significant
- common criterion “small chance” = 5% but sometimes 10%, 1% or some
other value is used
- to be an observed relationship must be statistically significant
- naturally some relationships will be labeled as statistically significant by mistake
- Two Warnings about Statistical Significance
- easier to rule out chance if observed relationship is based on very large #’s of
observations
- even minor relationship will achieve “statistical significance” if sample
is very large
- no necessarily a strong relationship or even one of practical importance
- very strong relationship won’t necessarily achieve “statistical significance” if the
sample is very small
- could mean that did not take enough measurements
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
How relationships between variables can be expressed quantitatively. Correlation = measures the strength of a certain type of relationship between two measurement variables. Regression = numerical method for trying to predict the value of one measurement variable from knowing the value of another one. Statistical relationship = natural variability exists in the relationship between the two measurements. Deterministic relationship = if we know the value of one, we can determine the value of the other exactly. Natural variability exists in relationship between two measurements in a statistical relationship. Statistical relationships useful for describing what happens to a pop. or aggregate. Stronger relationship -> more useful for predicting what will happen for an individual. Statistically significant = chances that a relationship that"s strong or stronger would have been observed in sample if there really were nothing going on in the population. To be an observed relationship must be statistically significant. Naturally some relationships will be labeled as statistically significant by mistake.
