About this calculator.

To test pairs of species to see whether the regression line for two variables measured in one species is consistently above the regression line for those variables measured in another species:

• enter your datasets (in the left window)
• enter SN1 SN2 SN3... the species name or ID for each measured organism
• enter X1 X2 X3 ... the measured X variable for each species
• enter Y1 Y2 Y3 ... the measured Y variable for each species
• enter the low and high X values for the test interval
• hit "Calculate!"

Each species must be represented by at least two individuals so that a regression line can be calculated.

Species names must not contain spaces or tabs.

Numbers in the input must be separated by spaces, tabs, or newlines.

Output consists of:

• regression coefficients of Y vs. X for each species
• for each pair of species:
• W1 is the height of the first species' regression line above the second species' regression line, at the left point of the interval (a negative value means the first is below the second)
• W2 is the height difference at the right point of the interval
• T1 (or T2) is W1 divided by the estimated standard deviation of W1 (or W2)
• the p value is the probability of Student's T being as large as or larger than the smaller of |T1| and |T2|, with df degrees of freedom. (df = n1 + n2 - 4, where species 1 has n1 points, and species 2 has n2) It is a one-tailed test. Note: the calculator prints a message when the two regression lines intersect in the interval, and does not do the test.

The output may look jagged, but has TAB-separated columns for easy copy/paste into a spreadsheet.

Warning: as with any all-pairs comparison, DON'T CONCLUDE ANYTHING without taking into account the overall experimental alpha. Even random datasets tend to produce "significant" outcomes if you perform all the pairwise tests! You should make a Bonferroni or Dunn-Sidak adjustment, reducing your comparison-wise alpha. You should strongly consider using multi-variate statistics, if you have several datasets.

(go back to the calculator)

The test is from:

R.K. Tsutakawa and J.E. Hewett, 1978. "Comparison of Two Regression Lines Over a Finite Interval". Biometrics 34:391-398.

The HTML, Java, and class files for this calculator are free.

This Java and HTML web page is by John Brzustowski. I appreciate any comments or criticisms. Although I strive to ensure the correctness of these programs, you use them entirely at your own risk. There is no warranty!