GLM Notation
How many parameter estimates are in an empty model?
One... b0 (the intercept/grand mean of the outcome).
What is SStotal referring to?
The sum of squared residuals for the EMPTY model.
The total variation in the outcome.
If you see .000 as the p-value output, how should you report this value?
p < .001
Because P-VALUES CAN NEVER BE ZERO
What function should you use to print mean, median, standard deviation, etc. of the variable you are interested in?
favstats()
What is the difference between b1 and b2 in the three group model?
b1 is the difference between the reference group (group 1) and group 2.
Likewise, b2 is the difference between the reference group and group 3.
Is it possible to have a PRE of 1.23?
No, PREs are proportions and are restricted to values between 0 and 1.
A researcher tests whether there is a difference in anxiety for individuals who underwent a mindfulness intervention vs. for those who did not. They fit a two-group model and find a p-value of .03.
Should we reject or fail to reject the null hypothesis that there is no difference between the control and treatment group (i.e., b1 = 0)?
Reject!
If you are interested in visualizing the relationship between a quantitative outcome and a predictor variable with 2 groups, what plots should you use?
Histogram, box plot, scatter plot/jitter plot
SECRET 200 POINT BONUS
What are the differences between shuffling and bootstrapping?
Shuffling randomly rearranges the labels or values in a dataset to break any real association, often to create a null distribution for hypothesis testing.
Bootstrapping resamples the data with replacement to estimate variability or build confidence intervals while preserving the structure of the original dataset.
What is the difference between the F-statistic and PRE?
The F-ratio shows how much better a model predicts the data compared to the leftover error while also adjusting for how many predictors the model uses, so more complex models must explain more variance to get a high F.
PRE simply tells you what fraction of the original error the model removes, without adjusting for model complexity.
Based on the 95% confidence intervals below, which estimate has a larger standard error?
(A) [.25, 2.75]
(B) [.75, 2.25]
A
If you fail to reject the null hypothesis when there is actually an effect, what type of error have yo made?
Type I vs. Type II error
Type II error
What is "e_i" referring to in GLM notation?
This refers to the residual value for an individual case, or the difference between the predicted value from a model and the actual observed value.
What is the F ratio? When comparing two models with the same outcome but different predictors, which model is better: the model with the higher or lower F ratio?
It is the ratio of error explained to error unexplained by our model. The model with the higher F ratio is preferred.
A memory experiment comparing two study strategies yields p = .12. What does this p-value say about how compatible the observed reduction is with the assumption that the intervention has no true effect?
This p-value indicates that the data is likely to have been generated under the "null" population.
What's the difference between a parameter estimate and a model prediction?
A parameter estimate describes the strength or form of the relationship that a model learns from data.
A model prediction uses those estimated parameters to generate an expected outcome for a specific input.