Lecture 11 Flashcards
Baseline groups
Comparison groups should be similar in terms of baseline characteristics.
Age, gender, ethnicity, variables of interest
Selecting a specific statistical test to perform requires some basic considerations about:
Type of
variable (independent vs dependent)
data (continuous vs discrete)
distribution (Normal/gaussian vs skewed)
Define variables
Representation of measures in an analysis
Can either be independent (the intervention) or dependent (the response)
Independent variable
The intervention. Defines the condition under which the dependent variable is measured. Experiementer controlled.
Independent variable affects change in the dependent variable.
The dependent variable
Response. Outcome or endpoint being measured or tested.
Continuous/infinite data
Infinite number of equally spaced values are possible. Measured. Cant count how many water drops are coming out of a hose, must measure it in L
Discrete/limited data
Limited number of values possible within the range of measurement. Counted, like single drops coming out of a faucet.
Defining data by scales
Ratio/interval scales
Ordinal scales
Nominal scales
Ratio or interval scales
Uniform intervals between consecutive measurements. Used to measure continuous data. Ex: 1 Km.
Ordinal scales
Rank of specific order where the interval values may not be known or constant. used to measure discrete data.
Ex: Leikert scale. Hot, hotter, hottest
Continuous data is measured by ____
Ratio/interval scales
Discrete data is measured by ____
Ordinal scales
Nominal scales
Used when data cannot be ordered, but values are discrete. Ex: age, race, gender. Named data.
1 dependent variable + no independent variable is what kind of analysis
Univariate analysis. Does not explain any relationship. Summarize, describe, look for patterns, normal distribution. NOT explanatory.
1 dependent variable + 1 independent variable is what kind of analysis
Bivariate analysis
Two variables. Compare.
1 dependent variable + 2 independent variables is what kind of analysis
Multivariate analysis
More than two variables. Evaluate relative importance.
Examples of univariate analysis
Mean, median, mode, range, variance/dispersion, standard deviation, bar charts, histogram, pie chart
Relationships evaluated during a bivariate analysis
Presence, strength, significant, descriptive or inferential.
Independent on X, Dependent on Y
What can you evaluate during a multivariate analysis
Relative importance.
Regression with a dependent variable: Linear (continuous) or logistic (dichotomous). Can model and form predictions. Strong statistical test.
Correlation. When no dependent variable: Relationship or connection. Not as strong as regression.
Difference between regression and correlation (both multivariate analysis)
Regression has a dependent variable. Strong statistical test.
Correlation does not have a dependent variable. Weaker statistical test.
n vs N
n samples
N entire population
Distribution of sample population measures
Central tendency.
Normal/gaussian distribution
Symmetrical, bell shaped distribution of values where the mean represents the highest point of the curve and whose spread is defined by the standard deviation.
Gaussian/normal curve.
__ represents the highest point of the curve
Spread is defined by ____
Mean, SD
Skewed distribution
Asymmetrical distribution of values. Look at tail to determine
Tail to the left = negative skew
Tail to the right = positive skew
How can you transform skewed values into a normal distribution
Taking the log, square root of the original value.
Types of quantitative tests
- Parametric.
Assume: same population normally distributed.
Allows comparisons between groups, typically based on sample means. - Non parametric.
Sample populations skewed. Comparisons typically made on sample medians.
What is a parametric test. What does it assume and what does it allow comparisons between
A type of quantitative test.
Assume: same population normally distributed.
Allows comparisons between groups, typically based on sample means.
What is a non parametric test. Comparisons are made based on
A type of quantitative test.
Sample populations skewed. Comparisons typically made on sample medians.
What is the first step after collecting data
Analyze data for a normal distribution.
Statistical test for sample sizes less than 50
Shapiro-Wilk
Statistical test for sample sizes more than 50
Kolmogorov smirnov
What does it mean about your data distribution if the P value is less than 0.05
It is not normally distributed
Do you use parametric or nonparametric tests to measure proportions?
Non parametric only. Chi square and McNemar’s tests. No parametric equivalent.
Do you use parametric or nonparametric tests to measure predictions?
Parametric tests only. Regression and multiple regression by least squares method. No non-parametric equivalent.
Why are parametric tests favored over non-parametric test?
The golden egg is to predict. You can only predict with parametric tests of regression.
Least squares
Type of parametric test. Algebraic procedure for fitting linear equations to data. Grew out of astronomy.
Goal of least squares
Minimize error of estimation
Why are parametric tests preferred over non-parametric tests
Parametric stats are considered more powerful
Continuous outcomes are generally more specific than categorical
Greater confidence is present for normally distributed data
Can generate predictions
