I am trying to assess the accuracy of 3 formulas (that estimate renal function vs actual measures of renal function. These are continuous measures however can also be binned into 5 categories (Normal, Mild, Moderate, Severe and Endstage). What test can be used in SPSS?
Hello Sian
you could use regression first and then dependent sample t test to see if there exists significant difference in the predicted value and actual value. Dependent sample t test should be not significant thus indicating that your predicted values were almost similar to actual values.
I hope it explains.
Good luck with your research!
I think you want to look at something like agreement. The classic approach here is the Bland-Altman plot: https://en.wikipedia.org/wiki/Bland–Altman_plot
SET UP. let f(x) be the estimated function or expected value. let g(y) be the actual function or observed values where f(x): [x1, x2, ..., xn] and g(y): [y1, y2, ..., yn]. We need to find the significance of the two sets: f(x) and g(x).
SIGNIFICANT DIFFERENCE FOR SAME SAMPLE SIZE: n1 = n2
Given that f(x): [x1, x2, ..., xn] and g(y): [y1, y2, ..., yn], the paired difference are:
(x1 - y1) = d1, (x2 - y2) = d2, ..., (xn - yn) = dn. Now work with the difference of the pairs:
d1
d2
...
dn
for the mean (d^) and standard deviation (Sd) of the difference. The test statistics to verify the significant difference (accuracy) is:
Td = d^ / Sd/sqrt(n)
SIGNIFICANT DIFFERENCE FOR DIFFERENT SAMPLE SIZE: n1 ≠ n2
Find the observed and expected means and standard deviations, then use the T test below:
T = (A - B) / Spool(Sqrt(1/n1 + 1/n2)
where A = (X1^ - X2^) and B = ( µ1 - µ2), assuming 1 is the standard or expected and 2 is the observed.
LINK:
Gopal Kanji, 100 Statistical tests
https://www.pdfdrive.com/100-statistical-tests-e97177.html
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS