For my master thesis I'm studying the relationship between Bitcoin's price index and several independent variables among which

**exchange trade volume**, amount of daily

**transactions**,

**sentiment**but also the price index of the

**Nasdaq**and

**Shanghai stock exchange**index. To examine this relationship, I use the ARDL model.

My problem is the specification of the variables. Using prices is not a good approach due to collinearity and inflated adjusted R2. Hence, I already transformed the dependent variable (BTC/USD) to log returns by: LN(Pt/Pt-1). It would be logical to do the same for the other two prices indices (

**nasdaq**and

**Shanghai**

**stock**

**exchange**). Regarding the rest of the variables (

**transactions**,

**exchange volume**), I thought the LN of the value would be sufficient: LN(X).

Turns out that the adj. R2 is not inflated anymore.

My questions are:

1. Is my approach correct does it correspond to what is normally done in statistical research using an ARDL model and

**prices/ returns**?

2. How should I interpret the coefficients? For instance, assume the coefficient of

**Nasdaq**is

**0.16**what does that mean (LN(return)? And assume

**exchange volume**is

**0.012**? LN(X).

3. I also have variables that are not transformed to logs. For instance, sentiment could not be transformed due to negative values. How do I interpret a coefficient of an independent variable that has not been transformed? Assume sentiment is 0.14 for instance.

Help is much appreciated!! Currently I'm really stuck due to the above questions....

Kind regards