15.5.2 Finding Y/X from X/Y Standard Curves

Video Image.png See more related video:Find X/Y Values after Fitting

The Find X/Y table allows you to obtain either a dependent variable value given an independent variable value, or an independent variable value given a dependent variable value, from the fit you performed on the data.

After fitting, there will a new worksheet named something like *FindXfromY* (or *FindYfromX*). Enter the Y (or X) values into the first column in the new sheet and the X (or Y) values will be calculated automatically.

Control over output options varies by fitting tool. In the Linear Fit tool, for instance, a check box turns on FindX/Y output and a second check box allows calculation of 95% LCL and UCL values at FindX/Y values. In the Nonlinear curve fit tool (NLFit), you can specify extra X/Y columns using a drop-down list. If there are multiple solutions, those values will be added in the extra columns. For example, in the model, there might be two x values for a given y value. So you might specify the Number of X Columns as 2 in the Find X from Y branch.

Generally, Find Y from X runs faster than Find X from Y. When finding Y from X, Origin uses the fit parameter values directly in the fitting model, to calculate the Y values. However, if you choose to find X from Y, Origin will not be able to derive an X~Y equation automatically. It has to calculate the approximate values by iteration. The algorithm used is illustrated below:

Origin first creates a uniform linear curve. For a given Y value, the range containing the Y value is found. For example, the Y value of a given point might fall into the range [yn, yn+1]. We know the X value should be within the range [xn, xn+1]. Then a new Y value y', which corresponds to X = (xn + xn+1) / 2, is computed. With the computed y', we can divide the range [yn, yn+1] into two subranges [yn, y'] and [y', yn+1]. Then y and y' are compared to see in which subrange y is located. These steps are repeated until the difference between y and y' is within some tolerance 1e^{-8}.

Note that the iteration method is used in Find X from Y. You can only find the X values for Y values within the source data range.

Algorithm for Calculating 95% Confidence Interval

When finding X from Y or Y from X, you can opt to calculate the 95% confidence interval.

  • Linear Fit
1. Find X From Y
Currently, Origin only supports calculating 95% confidence intervals for non-replicated data. For the specified Y value y_0, the resulting X is x_0.
If y_0 is within the range of measurement, by interpolation, the standard deviation s_{x_0} in x_0 is given by
s_{x_0} = \frac{s_\varepsilon}{b} \sqrt{1+\frac{1}{n}+\frac{\left(y_0-\bar{y}\right)^2}{b^2\sum_{i=1}^n \left(x_i-\bar{x}\right)^2}}
where b is the fitted slope, n is number of points, s_\varepsilon^2 is the sample variance, and \bar{x} and \bar{y} are the mean values of the X and Y data respectively.
If y_0 is out of the range of measurement, by extrapolation, the standard deviation s_{x_0} in x_0 is given by
s_{x_0} = \frac{s_\varepsilon}{b} \sqrt{\frac{1}{n}+\frac{\bar{y}^2}{b^2\sum_{i=1}^n \left(x_i-\bar{x}\right)^2}}
The interpolated or extrapolated X with (1-\alpha )\times 100\% confidence interval is calculated as:
[x_0 - t_{(\frac{\alpha}{2},n-2)}s_{x_0}, x_0 + t_{(\frac{\alpha}{2},n-2)}s_{x_0}]
where t_{(\frac{\alpha}{2},n-2)} is the critical t value for \alpha /2 (one tail) and n-2 degrees of freedom.
Note: If the weighting is specified, or there is a fixed parameter, the confidence interval for Find X From Y will not be calculated. If the Calculate 95% Confidence Interval box was checked, the result will be missing values.
2. Find Y From X
The confidence band will be used to calculate the confidence interval of Y.

  • Polynomial Fit and Nonlinear Fit
1. Find X From Y
The confidence band will be interpolated at a given Y to calculate the confidence interval of X. If X or Y is out of the range of input data, the confidence interval (CI) of X is not calculated.
2. Find Y From X
The confidence band will be used to calculate the confidence interval of Y.