Help with Statistics Data Analysis for Zero Inflated Negative Binomial
Budget: $50 – $0 USD
I need a statistician who can help me with a Generalized Linear Model for zero-inflated negative binomial data. I will provide the data and brief (here) but you will need your own access to software (e.g. R, STATA, etc.) to run the regression AND help me interpret the results.
What I'm trying to achieve is a score or an index that will summarise the performance of a store and chain of stores (the sum of scores for every store) based on data from Google reviews. In my mind, I imagine that I will take scores I construct for 4-5 Independent Variables and multiply the regression weights from explaining a Dependent Variable. I have done all the necessary work to have the data ready for you. The dataset contains 1100 cases.
From my limited knowledge & reading, I understand that I'm dealing with either a Zero Inflated Poisson or Zero Inflated Negative Binomial model because:
The Dependent Variable is the star rating (i.e. 1 to 5 star) for each of the 1100 reviews in the dataset.
The Independent Variables are:
1) Product score, which is derived from:
1.1) A count of the number of words that customers would use to describe their assessment of a retailer's Product offering (which includes no mention of any such words at all, hence the 'zero inflated'), multiplied by
1.2) A score of -1, 0, 0.5, 1 to represent the sentiment/favorability of the customer's assessment of the Product offering
2) Price score, which is derived from:
2.1) A count of the number of words that customers would use to describe their assessment of a retailer's Price offering, multiplied by
2.2) A score of -1, 0, 0.5, 1 to represent the sentiment
3) Service score, which is derived from:
3.1) A count of the number of words that customers would use to describe their assessment of a retailer's Service offering, multiplied by
3.2) A score of -1, 0, 0.5, 1 to represent the sentiment
I have both the raw and standardized scores for the 3 IVs, but as mentioned earlier, are both skewed and zero-inflated. Knowing enough not to bother with linear regression, BUT not knowing enough to understand or run the GLM (I don't have the software), I'm relying on you to:
1) Understand this brief and the nature of the data,
2) Run the regression with the required software,
3) Perform the necessary refinements to arrive at a satisfactory model,
4) Explain to me the procedures taken and the interpretation of the results.
What I'm trying to achieve is a score or an index that will summarise the performance of a store and chain of stores (the sum of scores for every store) based on data from Google reviews. In my mind, I imagine that I will take scores I construct for 4-5 Independent Variables and multiply the regression weights from explaining a Dependent Variable. I have done all the necessary work to have the data ready for you. The dataset contains 1100 cases.
From my limited knowledge & reading, I understand that I'm dealing with either a Zero Inflated Poisson or Zero Inflated Negative Binomial model because:
The Dependent Variable is the star rating (i.e. 1 to 5 star) for each of the 1100 reviews in the dataset.
The Independent Variables are:
1) Product score, which is derived from:
1.1) A count of the number of words that customers would use to describe their assessment of a retailer's Product offering (which includes no mention of any such words at all, hence the 'zero inflated'), multiplied by
1.2) A score of -1, 0, 0.5, 1 to represent the sentiment/favorability of the customer's assessment of the Product offering
2) Price score, which is derived from:
2.1) A count of the number of words that customers would use to describe their assessment of a retailer's Price offering, multiplied by
2.2) A score of -1, 0, 0.5, 1 to represent the sentiment
3) Service score, which is derived from:
3.1) A count of the number of words that customers would use to describe their assessment of a retailer's Service offering, multiplied by
3.2) A score of -1, 0, 0.5, 1 to represent the sentiment
I have both the raw and standardized scores for the 3 IVs, but as mentioned earlier, are both skewed and zero-inflated. Knowing enough not to bother with linear regression, BUT not knowing enough to understand or run the GLM (I don't have the software), I'm relying on you to:
1) Understand this brief and the nature of the data,
2) Run the regression with the required software,
3) Perform the necessary refinements to arrive at a satisfactory model,
4) Explain to me the procedures taken and the interpretation of the results.