Questions tagged [interaction]
A situation where the effect of an explanatory variable may depend on the value of another explanatory variable.
2,811 questions
3
votes
1
answer
24
views
How to handle calendar year as a continuous predictor with a mismatched train/test time horizon?
I am using Ordinal Semiparametric Regression (Frank Harrell's rms package) to model overall survival in patients with brain tumor.
My training data is from the SEER database (covering years 2004 to ...
- 171
7
votes
2
answers
275
views
How to decide if an interaction exists: graphically/interaction terms/contrasts of slopes
I have fit interaction models of the form: phenotype ~ genotype * environment, based on theory.
I am assigning environment (GFR, in this case) as the moderator.
I have three scenarios:
A: Non-parallel ...
- 191
4
votes
1
answer
56
views
How to test interactions involving latent profiles
I'm using R but this is fundamentally a conceptual/statistics question so I'm posting here (apologies if it's not suitable and I will delete/repost to Stack ...
- 43
0
votes
0
answers
55
views
Conceptual issues with using compositional data in model interaction term?
Can compositional data (seagrass % cover of multiple species - each value is mean of 10 1m^2 area per site) be used in an interaction term like this (and are GAMs somehow better equipped to handle ...
- 2,631
5
votes
1
answer
179
views
How to explain the collider bias from treating a post randomization variable as a moderator?
I have been struggling with explaining and understanding the impact of collider bias from intuitive standpoint particularly when using a post randomization variable as a moderator in a longitudinal ...
- 1,126
4
votes
2
answers
127
views
Should I keep the interaction test in this comorbidity stigma study?
I am running a 4-group vignette study: Control, Condition A, Condition B, and Comorbid A+B.
I chose this design because A+B is common in real life, and I wanted to test whether the comorbid ...
2
votes
0
answers
29
views
Interactions in time series
In time series, is it possible to have an interaction between an AR and a MA term?
For example an ARMA(1,1) model:
Standard (without interaction): y_t = A * y_{t-1} + B * e_{t-1} + e_t
With ...
user506655
2
votes
1
answer
83
views
Hierarchical regression for moderation: Is simple regression acceptable in Models 1 & 2?
The Problem
The same path (Platform Interactivity → Trust) shows completely opposite results:
In SEM (controlling for all predictors):
Coefficient: β = -0.035
Significance: p = 0.380 (not significant)...
1
vote
0
answers
26
views
With longitudinal data, does using GEE over time-stratified linear regression increase effective sample size?
I am looking for predictors of subjective impairment (questionnaire with score from 0-40) in a cohort of ~300 patients. Each of the patients have 1-7 visits over the last few years at fixed intervals.
...
- 549
5
votes
1
answer
79
views
GAM with repeated measure outcome: pvalue to test `Group*Time` interaction or any Group effect at all?
Using GAMs to model a longitudinal repeated measure outcome has been reviewed here and here.
However, I have not found an answer for how to calculate pvalues for testing the ...
- 1,005
5
votes
4
answers
278
views
How to specify contrast for interaction in mixed anova with 3 conditions and 2 groups
20 people are divided into two groups (between subject variable) - 12 and 8 members in each. Each person is measured under 3 conditions (within subject variable).
I am using means model ...
- 4,030
3
votes
2
answers
180
views
Multinomial models interaction on probability
I have been experimenting with brms over the past few days, and there is something I can’t quite figure out concerning the approach with logistic regression and log-odds to probability transformations....
- 41
2
votes
1
answer
49
views
Can decision trees spot XOR patterns in the presence of noise?
I have been investigating how the CART algorithm (specifically scikit-learn's implementation) handles interaction terms like the XOR problem. I’ve noticed a behavior where adding noise in the form of ...
- 21
6
votes
1
answer
106
views
Two-Way ANOVA with Missing Observations: Type I vs. Type II Sums of Squares
I am currently writing my bachelor’s thesis and have only a limited background in statistics.
I have four groups with four observations each, except for one group which contains only three ...
- 415
4
votes
1
answer
415
views
Main effect becomes insignificant when adding a moderator, but interaction is significant. Not sure how to interpret
I have a continuous independent variable which ranges from -5 to 5, and a continuous dependent variable which was measured with a Likert Scale ranging from 1 to 7. My main effect with just these two ...
- 43