HypothesisSelection.Rmd

---
title: "HypothesisSelection"
output: html_document
date: "2025-05-28"
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

## Is Model 7 Right For Your Experiment?

Model 7 can be used to analyze data from any kind of mass spectrometry experiment, however it is especially useful for data with variability across multiple levels of grouping factors (i.e. translational proteomics). This is because it is based on linear mixed-effect models that can account for both fixed and random effects. In contrast, t-tests has no way of accounting for random effects. T-tests are appropriate when there are no random effects to account for (flat data). In these cases, Model 7 will still work, but it's unnecessarily sophisticated. In short: if your data has random effects / multiple levels of grouping factors, use Model 7. If not, use a simpler model like a t-test.

Additionally, Model 7 is only suited for comparison of two groups, A and B. If a third group C needs to be compared, a model such as mixed-effects ANOVA would be more suitable. (This section probably needs to be improved for clarity, but I'm not sure how to.)

## Choosing The Right Hypothesis

To preserve the random effects of hierarchical data, special care must be taken to avoid reducing the dimensionality of (flattening) the data. Model 7 accomplishes this by forcing you to choose between 3 distinct hypothesis types that compare different aspects of hierarchical data. In doing so, it also (maybe?) avoids mixed effect modeling (if so, what's the benefit?).

*Question*: I've only seen examples with conditions and cell types as the two sources of variability. How does Model 7 deal with, say, cell type and patient and hospital levels at the same time? Seeing an example would be super helpful.
- Partial answer: the BioReplicate column in the dataframe accounts for patient-level variability. Maybe patient and cell type are the only random effects we want to account for at the moment?

### Hypothesis 1 - Comparing conditions within the same cell type  
For 2 cell types and 2 conditions, there are 2 versions of hypothesis 1:  
- Condition 1 vs Condition 2 within cell type 1   
- Condition 1 vs condition 2 within cell type 2  

For example, we would use hypothesis 1 if we wanted to compare protein expression in smooth muscle cells between wildtype and diseased subjects.

### Hypothesis 2 - Comparing cell types within the same condition
For 2 cell types and 2 conditions, there are 2 versions of hypothesis 2:
- cell type 1 vs cell type 2 within Condition 1
- cell type 1 vs cell type 2 within Condition 2

For example, we would use hypothesis 2 if we wanted to compare protein expression between smooth muscle cells and fibrous cells within wildtype subjects.  

### Hypothesis 3 - Difference of Differences
Hypothesis 3 effectively compares the difference between version 1 and version 2 of Hypothesis 1.  

Keeping to the same dataset as the previous example, hypothesis 3 would be comparing the difference in protein expression in smooth muscle cells between wildtype and diseased subjects with the difference in protein expression in fibrous cells between wildtype and diseased subjects. 

This can be expressed as (smc X WT - smc X diseased) - (fibrous X WT - smc X diseased)