Data Analysis Section - Githubissues

9yusup4 commented 4 months ago

I hope this message finds you well. I am writing to seek your expertise regarding a significant issue I've encountered in my data analysis work. Your guidance would be immensely appreciated.

Following the prescribed steps, I completed a conditional logit model. Subsequently, I proceeded with willingness to pay (WTP) analysis. However, to my surprise, the results of WTP revealed p-values greater than 1. Upon scrutinizing the data, I ensured that all procedures were meticulously followed. Despite this, there appears to be a discrepancy between the results of the graphical representation and those of the conditional logit model.

I am reaching out to you in the hopes of discussing this matter and identifying the root cause of the problem. Your insights and expertise would be invaluable in resolving this issue.

Attached are the outputs of the conditional logit model, WTP analysis, and the graphical representation for your reference.

Thank you for your time and consideration. I eagerly await your response.

danielpereztr commented 4 months ago

Hi.

Thank you for pointing out this issue.

I see several things here that are worth noting:

First, the cont_price coefficient is not significantly different from zero. This might be a coding error (when you inputted the numeric values of the price variable creating the continuous variable) or it might be a result. I say this because that coefficient is usually interpreted as "the amount of utility that the respondent gets (or loses when it's negative) when they spend a monetary unit (for example, an euro or a yuan depending on how you expressed your prices). So, since your coefficient is zero (not significantly different from zero), it means that no utility is lost when spending money, which is a bit odd. This might happen because respondents did not take the price into account, so a solution might be increasing the price gap between levels (but this is only a speculation, don't blindly trust what I say).
As a result of what I mentioned, most WTP (willingness to pay) values are negative and not significant. In this scenario, you cannot trust these estimates.
There's definitely something wrong going on with the graph. I don't know if it's caused by what I mentioned before or if there is something else. I'll have a look, but for now, only trust the numbers.

I hope this helps you.

9yusup4 commented 3 months ago

I am reaching out once more regarding the discrepancies in my data analysis, for which you previously offered valuable insights.Following your observations, I am curious about the next steps I should undertake to rectify the situation. Specifically, could you elaborate on how I might adjust the price gap between levels to address the non-significant cont_price coefficient? Your further guidance on this initial concern would be incredibly helpful.

danielpereztr commented 3 months ago

Hi.

You can try by increasing your sample size (it's normal to have no significant coefficients with a small sample size) or running again the DCE with a higher difference among the price levels.

Best!

danielpereztr / DCEtool

Data Analysis Section #7