Linear Regression modeling on Graphics Cards: Am I missing something?

[Engineiro]

:wave:
I am trying to understand how to choose a graphics card based on a price to performance value for each card.

I have a number of price/performance values for a several graphics cards and I plotted them. I naturally thought that it may be best to find a linear regression line for this data set and figure out their price to performance values from the delta between the linear regression line and the datapoint.

But then I thought, I may be overcomplicating this. I can simply divide the performance by the price and look where performance is high and price is low.

Am I thinking along the right lines? The way I understand linear regressions are used to estimate or predict the value of an dependent (price) variable based on the value of an independent (performance) variable, and although this may be useful to me in the future this is not necessary for the analysis I am doing now to find the price performance leaders among a dataset.

Thanks for considering this question and dealing with my noobiness.

 

[BGM]

Some how you want to do a cost effectiveness analysis on the product.

Regression analysis is a powerful tool to discover the association/relationship between the variables. One thing to remind you here is that a linear relationship may not fit the data well. It depends on your measurement.

Also, one thing that the regression model can do is interpolation/extrapolation. For example, let say you have already estimated the model parameters and the model fitting the data well. Now you look at a new product - and you can give an expected performance with a given price; or if you are a manufacturer, some how you can know how to set the corresponding price based on the performance of your product.

 

[Engineiro]

Some how you want to do a cost effectiveness analysis on the product.

Regression analysis is a powerful tool to discover the association/relationship between the variables. One thing to remind you here is that a linear relationship may not fit the data well. It depends on your measurement.

Also, one thing that the regression model can do is interpolation/extrapolation. For example, let say you have already estimated the model parameters and the model fitting the data well. Now you look at a new product - and you can give an expected performance with a given price; or if you are a manufacturer, some how you can know how to set the corresponding price based on the performance of your product.

The datapoints do show a pattern in their behavior i.e. the higher the price the higher performance you can expect. So it sounds like I may want to use regression at some point but for now I may be able to do a simple ratio comparison?