Like they are on the same axis of some specific effect or result so they are somewhat statistically meaningful in determining the range of outputs or responses
Is orthogonality usually of a gradient or range like nature, is that essentially what is implied by orthogonality and the orthogonal items being on the same axis?
Edit- i think i might have misunderstood orthogonal
It’s been years since I’ve been in the lab but it really will depend a lot on the subject matter and the type of experiment.
If it’s a subject matter that is fairly well explored and defined, the alternative hypotheses might be fairly straightforward. Take, for example, an experiment from a while ago where entomologists suspected that desert ants navigate by using dead reckoning, effectively counting their steps, remembering their changes in direction measured by a biological compass, and integrating them together, in a process similar to “fusion” in electronic position sensors.
To validate part of this hypothesis, they needed to get more granular and isolate one part of it. So, they formulated a “sub-hypothesis” that stated that the ants had some sort of innate awareness of the distance that they covered with each step, knowing the length of their legs and this their stride length, similar to how cats know their healthy body width. The experimental hypothesis would be something like:
“Altering the length of desert ant legs will result in navigation failure with longer legs causing them to overshoot and shorter legs causing them to undershoot. The navigational trajectories should otherwise be identical.”
Building alternative hypotheses for this relatively simple experiment, prior to conducting it would be straightforward, as you appear to be suspecting. They could be as simple as:
“The length of the desert ant’s legs will have no impact on their navigation because they are not directly related. This will be apparent through the ants showing no discernable difference in the paths that they take when navigating, regardless of leg length.”
“The length of the desert ant’s legs will have some impact on their navigation but, they are able to compensate for discrepancies in stride length through some as of yet unknown mechanism. This will likely be apparent in statistically significant distance-related navigation errors in their paths.”
After the experiment, the data would be analyzed and checked for a match against the established hypotheses. If there is not a good match or there is an unexpected shape to the data, further experiments may be required to see if it is an anomaly or if something else might be going on.
(In this case, it was found that, yes, desert ants have some sort of innate awareness of what their stride length should be and changes in their leg lengths throw off their navigation, as expected.)
Now, when it gets to subjects that are less clear and established, alternative hypotheses can get a lot more challenging because often the difference between the data fit that proves or disproves a hypothesis can be miniscule. Or, the data points might form a completely unexpected shape that doesn’t match currently known phenomena.
I hadn’t heard “antihypothesis” before, but there is a whole statistical topic of hypothesis testing.
https://en.wikipedia.org/wiki/Statistical_hypothesis_test
I guess you are asking about “alternative hypothesis”:
https://en.wikipedia.org/wiki/Alternative_hypothesis
The stuff there might help. I’m used to the maybe simplified example of a drug trial, where the null hypothesis for an anti-baldness drug is that it’s equivalent to placebo.
I guess you are asking about “alternative hypothesis”:
Tbh I thought they were talking about the null hypothesis.
I very well could be, I’ve seen it in the wild somewhere but i cant point to it
I was taught that it was called the antihypothesis in school
Ironically i dont think they ever covered that far and it was simply
- Purpose
- Hypothesis (like 1 singular hypothesis, not hypotheses)
- Materials
- Process
- Results
- Analysis
Wondering if my all university/academic level science classes were less than exhaustive relative to what they should have been altho i cant complain
Wouldnt it be better to have all three? I feel like if you only do null and hypothesis you miss out on something seemingly opposite while still orthogonal or maybe thats nonsensical and I’m just too tired right now haha 🤪
Hypothesis testing is not of the form “what causes this?” What you’re suggesting seems to be along those lines.
Instead it’s more “can we say with high confidence that this specific factor causes this?”
That doesn’t mean you can’t test other factors! You can test them all with enough time and resources.
There are multi-factor statistical tools like ANOVA. But they still depend on you identifying what the factors might be.
But if you have factors A, B, C, and D in your analysis, and it’s actually the totally unknown factor E… you might find a lot of unexplained variance in your statistics, or you might mislead yourself into thinking it’s ABCD and never discover what E actually was.
But at the end of the day that’s just a fancier form of “does this specific thing cause this effect.”
And the essence of science is discovering Factor E and testing it, with new hypotheses.
