Hello everyone, I've come across a complication while doing some data analysis and was looking for some advice and interpretations.

I need to a run a two way ANCOVA on some data; I'm testing the effects and interactions from 2 independent variables, and our experiments were performed on samples that came in multiple batches so values representing untreated samples per batch need to be included as a co-variate in order to account for batch variation.

I've needed to run these models for analysis of different metabolites in the samples. But there is significant heteroscedasticity and heterogeneity of variance in multiple analyses. I have tried transforming values, but this hasn't worked ( in some cases this even violated normal distribution).

I have found that if there are heterogenous variances in your models then you can either: carry out a two way ANCOVA with robust standard errors (HC3 or HC4) , or weighted least squares regression. However, there are no specific details on how to perform these in SPSS, or how to interpret any outputs.

I am certainly not a statistician, therefore I'd like to run my queries by the community so I can determine the best way forward. Therefore, the main questions I have are as follows:

1. How do you perform a two way ANCOVA with robust standard errors in SPSS version 25?

2. What exactly does the robust standard error accomplish; does it provide homogeneity to your dataset or does it just give confidence intervals and outputs that are appropriate for heterogenous data? (again, I'm no statistician so I'm not very familiar with these terms)

3. After applying robust standard errors, can I still use the 2 way ANCOVA to report my data? Or will a non-parametric test need to be used?

Any help or advice would be massively appreciated, thank you in advance for your attention.