Hi guys,
If I add different terms in a model for factorial ANOVA, does it mean I adjust each of them to the rest?
For example in Y ~ A + B + C model, the terms (factors) A, B, and C are adjusted by each other, correct?
Thanks!
factor adjustment in ANOVA
Re: factor adjustment in ANOVA
no entirely sure what you mean by "adjust each of them to the rest" ... but we do expect adding more factors into a model will impact the results for the preexisting factors.
jonathon
jonathon
Re: factor adjustment in ANOVA
The answer is yes and no.
When the factors are not correlated, i.e., an additive model, adding more factors does not adjust or change the effect of existing factors. So, the answer is no.
When the factors are correlated, the effect size of each factor will be changed. So, the answer is yes in this case.
When the factors are not correlated, i.e., an additive model, adding more factors does not adjust or change the effect of existing factors. So, the answer is no.
When the factors are correlated, the effect size of each factor will be changed. So, the answer is yes in this case.

 Posts: 15
 Joined: Mon Sep 26, 2022 8:41 am
Re: factor adjustment in ANOVA
deleted post
Last edited by jamovi_user on Wed Feb 15, 2023 3:15 pm, edited 2 times in total.

 Posts: 15
 Joined: Mon Sep 26, 2022 8:41 am
Re: factor adjustment in ANOVA
deleted post
Last edited by jamovi_user on Wed Feb 15, 2023 3:15 pm, edited 3 times in total.

 Posts: 15
 Joined: Mon Sep 26, 2022 8:41 am
Re: factor adjustment in ANOVA
Thanks for the replies!
I thought that adjustment (i.e. adding additional factors in my naive thinking)
should be used at least when the number of samples in different groups is (rather) different.
I thought it would help to clarify whether the effect comes from a factor B rather
then say the low number of subjects in factor A.
For example, in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2276914/ paper the authors write:
"After adjusting for confounders, male sex was associated positively with ..."
and then:
"Linear regression was employed for multivariate analyses with serum immunoglobulin levels as dependent variables. For covariates, age (in years) entered the equation as a quantitative variable, and binary variables entered the equation as ‘1’ (‘present’ or ‘yes’) or ‘0’ (‘absent’ or ‘no’)... Variables were forced to enter the equation in all models. To account for the stratified sampling, a designbased analysis including compensatory weights was performed for the estimation of immunoglobulin levels in the overall population."
And here: https://phenome.jax.org/measures/41602 both adjusted and unadjusted means are listed:
Also here: https://pubmed.ncbi.nlm.nih.gov/12580912/
Alcohol consumption of more than 14 units/week was associated with an increase in serum IgE levels
after adjusting for age, gender, allergic sensitization and smoking
Also here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9125820/ (table 2):
"Multivariate logistic regression analyses were performed to determine independent correlates of depressive symptoms. All these were adjusted for in three models: Model 1: no adjustment; Model 2: adjusted for age, sex, and ethnicity; Model 3: age, sex, ethnicity, centenarians, systolic blood pressure (SBP), diastolic blood pressure (DBP), RBC, hemoglobin, MCHC, WBC, neutrophil, CRP, immunoglobulin A, immunoglobulin G, immunoglobulin M, immunoglobulin E, complement C3, and complement C4. "
Also here: https://pubmed.ncbi.nlm.nih.gov/35606877/
"After adjusting for all covariates, we found that immunoglobulin A levels were positively associated with depression."
Maybe it is just about ANCOVA...?
I thought that adjustment (i.e. adding additional factors in my naive thinking)
should be used at least when the number of samples in different groups is (rather) different.
I thought it would help to clarify whether the effect comes from a factor B rather
then say the low number of subjects in factor A.
For example, in https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2276914/ paper the authors write:
"After adjusting for confounders, male sex was associated positively with ..."
and then:
"Linear regression was employed for multivariate analyses with serum immunoglobulin levels as dependent variables. For covariates, age (in years) entered the equation as a quantitative variable, and binary variables entered the equation as ‘1’ (‘present’ or ‘yes’) or ‘0’ (‘absent’ or ‘no’)... Variables were forced to enter the equation in all models. To account for the stratified sampling, a designbased analysis including compensatory weights was performed for the estimation of immunoglobulin levels in the overall population."
And here: https://phenome.jax.org/measures/41602 both adjusted and unadjusted means are listed:
Also here: https://pubmed.ncbi.nlm.nih.gov/12580912/
Alcohol consumption of more than 14 units/week was associated with an increase in serum IgE levels
after adjusting for age, gender, allergic sensitization and smoking
Also here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9125820/ (table 2):
"Multivariate logistic regression analyses were performed to determine independent correlates of depressive symptoms. All these were adjusted for in three models: Model 1: no adjustment; Model 2: adjusted for age, sex, and ethnicity; Model 3: age, sex, ethnicity, centenarians, systolic blood pressure (SBP), diastolic blood pressure (DBP), RBC, hemoglobin, MCHC, WBC, neutrophil, CRP, immunoglobulin A, immunoglobulin G, immunoglobulin M, immunoglobulin E, complement C3, and complement C4. "
Also here: https://pubmed.ncbi.nlm.nih.gov/35606877/
"After adjusting for all covariates, we found that immunoglobulin A levels were positively associated with depression."
Maybe it is just about ANCOVA...?