The Beast of Bias + Mediation

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/19

encourage image

There's no tags or description

Looks like no tags are added yet.

20 Terms

1
New cards

How do we detect outliers? (2)

1. Graphs, histograms, scatterplots (less helpful with several predictors) 

2.       Standardised residuals  

<p><span>1.</span><span style="font-family: Times New Roman; line-height: normal; font-size: 7pt;">&nbsp;</span><span>Graphs, histograms, scatterplots (less helpful with several predictors)&nbsp;</span></p><p class="MsoNormal"><span>2.</span><span style="font-family: Times New Roman; line-height: normal; font-size: 7pt;">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; </span><span>Standardised residuals&nbsp;&nbsp;</span></p>
2
New cards

Residual

The difference between an actual observed value and the value predicted by a model.

3
New cards

Standardised residuals

Residuals expressed in standard deviation units.

~ 95% should fall within ±2 and 99% within ±2.5. Absolute values ≥3 suggest outliers.

4
New cards

Influential cases

Observations that substantially affect the fitted linear regression model (e.g., its slope or intercept).


An outlier is not necessarily influential.

5
New cards

Methods of detecting influential cases

DF beta

Cook’s distance

6
New cards

DF beta

Measures how much an individual regression coefficient (b) changes when an observation is removed. Absolute standardised values >1 are concerning.

<p>Measures how much an individual regression coefficient (b) changes when an observation is removed. Absolute standardised values &gt;1 are concerning.</p>
7
New cards
<p>Cook’s distance</p>

Cook’s distance

Measures the influence of a single case on the model as a whole.

Absolute values >1 are cause for concern, any larger than 0.5 are likely to be influential. 

<p><span>Measures the influence of a single case on the model as a whole.</span></p><p class="MsoNormal"><span>Absolute values &gt;1 are cause for concern, any <strong>larger than 0.5 are likely to be influential.&nbsp;</strong></span></p>
8
New cards

Key assumptions of the General Linear Model (GLM)

Linearity (straight-line relationships)

Additivity (predictor effects add together)

Spherical errors (homoscedastic + independent errors)

9
New cards
<p>Additivity</p>

Additivity

The combined effect of predictors equals the sum of their individual effects

10
New cards

Spherical errors

Errors must be independent (uncorrelated) and homoscedastic (constant variance).

11
New cards

Errors vs Residuals

Errors: The true, unobservable difference between an actual outcome and the real population prediction.

Residuals: The observable difference between an actual outcome and your sample model's prediction

12
New cards

Autocorrelation

Autocorrelation happens when errors are correlated with each other (often across time or sequence). It violates the independence assumption.

(That errors for different data points must not be linked or predictable from one another)

13
New cards

Homoscedasticity vs heteroscedasticity

Homoscedasticity: The error (spread of data) stays the same across all values.


Heteroscedasticity: The error (spread of data) changes as values get bigger or smaller.

<p><strong>Homo</strong>scedasticity: The error (spread of data) stays the <strong>same</strong> across all values.</p><p></p><p><strong>Hetero</strong>scedasticity: The error (spread of data) <strong>changes</strong> as values get bigger or smaller.</p>
14
New cards

Central Limit Theorem (CLT)

The more samples you have, the distribution will closely resemble a normal disturbution regardless of the original population distribution. 

15
New cards

Sample size and normality

Large samples generally reduce concerns about non-normality for inference, but do not guarantee normality. In smaller samples, bootstrapping may provide an alternative for estimating uncertainty.

16
New cards

Robust model procedures

knowt flashcard image
17
New cards
<p>Bootstrapping</p>

Bootstrapping

Used for small samples

Repeatedly resamples observations WITH replacement, recalculating a statistic each time to create an empirical sampling distribution.


picks a number and puts it back in and creates a mean from this. and again and again

18
New cards

Heteroscedasticity-consistent (HC) standard errors

Robust standard errors that adjust for non-constant error variance, improving statistical inference when homoscedasticity is violated. Calculated using a sandwich estimator.

HC3 and HC4 models work best

19
New cards

Mediation

Meditation is when the relationship between a predictor and outcome can be explained by their relationship to a third variable.

<p>Meditation is when the relationship between a predictor and outcome can be explained by their relationship to a<strong> third variable.</strong></p>
20
New cards

The total effect between a predictor and outcome can be broken down into:

The direct effect: The relationship between the predictor and outcome adjusting for the mediator (path c)

The indirect effect: The relationship between the predictor and outcome via for the mediator (path ab)


Total effect = c + (a x b)

<p>The <strong><u>direct effect</u></strong>: The relationship between the predictor and outcome <strong>adjusting</strong> for the mediator <strong>(path c)</strong></p><p class="MsoListParagraphCxSpLast">The <strong><u>indirect effect</u></strong>: The relationship between the predictor and outcome <strong>via</strong> for the mediator <strong>(path ab)</strong></p><p class="MsoListParagraphCxSpLast"></p><p class="MsoNormal" style="text-align: center;"><strong><mark data-color="red" style="background-color: red; color: inherit;">Total effect = </mark></strong><mark data-color="red" style="background-color: red; color: inherit;">c + (a x b)</mark></p>