Toggling Midpoint for Interpolation
Posted: Mon Jan 18, 2016 12:11 pm
Hey,
This is more of a math question, but has some cool implications for certain econometric models.
Consider a probit model, where certain predicted outcomes of the model are of interest, for example when y_hat = .5
When plotting the x-axis of a response probability curve by interpolating the range of the desired regressor as follows:
(for the general case, don't use my max and mins)
I have found that certain key points, like .5, are not guaranteed to be points. My latest program successfully created the graph, but it only hit the points: .48 and .54. Is there a way to interpolate so that we go from min to max as normal, but also hit a mid point (say .5) for clarity/convenience?
I guess this is might not be feasible, seeing as we can't know what y_hat will be until after evaluating the cnorm of x + other variables at their means, or can we? If not, can eviews maybe interpolate, or solve for y_hat = .5 with my existing coordinate pairs? I'd prefer not using the model approach, if possible.
Thank you
This is more of a math question, but has some cool implications for certain econometric models.
Consider a probit model, where certain predicted outcomes of the model are of interest, for example when y_hat = .5
When plotting the x-axis of a response probability curve by interpolating the range of the desired regressor as follows:
Code: Select all
series x = 5+12*(@trend-@min(@trend))/(@obs(@trend)-1)I have found that certain key points, like .5, are not guaranteed to be points. My latest program successfully created the graph, but it only hit the points: .48 and .54. Is there a way to interpolate so that we go from min to max as normal, but also hit a mid point (say .5) for clarity/convenience?
I guess this is might not be feasible, seeing as we can't know what y_hat will be until after evaluating the cnorm of x + other variables at their means, or can we? If not, can eviews maybe interpolate, or solve for y_hat = .5 with my existing coordinate pairs? I'd prefer not using the model approach, if possible.
Thank you