How to use t.test {stats} when comparing a mean against a sum (NOT a mean against a mean) ?

[cwright1]

I'm trying to use R's built-in stats package to statistically measure the difference between what I observe and what I expect. I believe the best test to use is a t-test (using t.test) - however that seems to just compare means. Here's my simple data:

brightness_1 <- c(1,1,2,2)
brightness_2 <- c(8,9,8,10)

brightness_combined <- c(70, 71, 71, 74)

Let's say I take a lightbulb and put some energy into it, and measure the brightness. I do it four times and the results are brightness_1. Then I increase the energy and measure four more times - that's brightness_2.

Okay - so if I combine however much energy I used for both experiments, I would expect to see the brightness is the combination. Maybe I'd expect something like an average of ~10. However, what I see is that my values are much higher than expected : brightness_combined.

How can I measure this? If I use a t-test with the average of brightness_combined set as the 'mu' value, it will compare the means....but I want it to compare the sum of brightness_1 and brightness_2. If I were to sum them beforehand, I lose the number of samples, right?

[edgarbenitez]

You must to evaluate the energy like a continuous variable instead a categorical one... You must evaluate an exponential model or a polynomic one (second level maybe)...

[cwright1]

Thank you for your input. How can I do this with my data?

[edgarbenitez]

Don you have the exact values of energy? Wats, joules, amps?

[nirgrahamuk]

example of lm with exponential terms.
The approach is to build a model of the phenomenon and then evaluate that model.

library(tidyverse)

(mydata <- tibble(
  brightness_1=c(1,1,2,2),
  brightness_2=c(8,9,8,10),
  
  brightness_combined=c(70, 71, 71, 74)
))

(mydata2 <- mutate(mydata,
                
  exp_b1 =exp(brightness_1),
  exp_b2 =exp(brightness_2)
))
lm1 <- lm(brightness_combined  ~exp_b1 + exp_b2,
          data = mydata2)

mydata2$lm_pred <- predict(lm1,newdata = mydata2,type="response")

summary(lm1)
mydata2

[cwright1]

I wasn't familiar with the exp feature before, thank you for the example!

However, isn't this ignoring that the 'expected' is supposed to be brightness_1 + brightness_2 ? In my own experiment I expect the combined_brightness to be the average of (1,1,2,2) + average of (8,9,8,10) - meaning my actual observation (70, 71, 71, 74) is much higher compared to that sum. But I think here if I'm correct it's the actual observation compared to the exponential model of each drug?