Junaid -

Are you saying that you have the information to plot any point (xi, predicted yi)?  Can you just plot predicted y on the x-axis, and x on the y-axis to obtain what you want? 

Do you have information on estimated variances of the prediction errors, so you could similarly 'flip' those bounds on a scatterplot graph? 

Does this relate to your question and/or serve your purpose at all? 

You might get into some definitional difficulties by changing the roles of x and y, by the way.  I mean, you are changing which is the dependent variable. That is a subject matter (e.g., economics, or biology context) argument, as the statistics are not for purposes of determining causality. That is a theoretical - subject matter consideration. 

Cheers - Jim  

PS -

Estimated residuals often approximately sum to zero in a given example application.  For some models they always exactly sum to zero: see

https://www.researchgate.net/publication/261474011_The_Classical_Ratio_Estimator_%28Model-Based%29

there is a section in that paper which is titled "Conditions Under Which Estimated Residuals Sum to Zero." 

For a regression model such as

 y = f (x) + e 

f(x) is the "predicted" y. 

So if the above approach works for whatever you are doing, then from your regression equation, predicted y is f(x), not the observed y-values that were used to obtain the regression originally. 

- If you have a regression plotted already, and the above is what you want, then you already have (x, predicted y) plotted, so in that case, just reverse the axes. Your software may do that for you.

Article The Classical Ratio Estimator (Model-Based)

Dear Sir James R Knaub,

Yes Sir i can plot any point (xi, predicted yi) by using the least square regression equation but i don't have any information on estimated variances of the prediction errors, only i know more is the value of R (square). So i think i can not use the equation y = f (x) + e.

Can you please elaborate that how can I find the inverse function of least square regression equation from the Graph or Equation without knowing the original data

                                                                   OR

how can I get the original data of scatter plot by using the least square regression equation and value of R(square) ?