Why am I no longer an anarchist

[teo123]

IF ships disappearing from the bottom were literally the only reason we suspected the Earth to be round, then challenges to that would surely be taken more seriously as they probably were in the era of that discovery.

By that logic, if the k-r pattern in the Croatian river names were the only reason I suspected the Illyrian language to have been a centum language, the challenges to that should be taken seriously. But the k-r pattern is not the only reason I think Illyrian was a centum language. Here I list five reasons to think that. Is that how it works?
And what do you think of the neo-scholastics saying that it's true that each of the Aquinas'es five arguments for the existence of God is weak, but that together they make a rather strong case for the existence of God?

[teo123]

Doing a survey of *criminals* would be another approach, find out how many have been frightened off from victimizing somebody because the person had a gun.

But wouldn't such a study suffer from the exact same problems that Gary-Kleck-like studies do? The (supposed) problems with the Gary-Kleck-like studies are that people are massively telescoping DGU events that really happened and are making up the events. Criminals would also be expected to massively telescope real DGUs and make up some DGUs that didn't really happen, right?

[teo123]

@Red, I am interested, what do you think about that new argument I found on a Croatian-language historical revisionist website for Vukovar Denialism? The argument goes that Vukovar Massacre is an extraordinary claim, much more so than the Holocaust is, because we are living in the most peaceful time in human history. Even if it is reasonable to believe there was a Holocaust in the 1940s, that doesn't mean it is reasonable to believe there was a Vukovar Massacre in 1991.
The premise of that argument is almost certainly true, because there are entire books written about how this is the most peaceful time in human history and what might be the causes of that. Steven Pinker's "The Better Angels of our Nature" is one such book. You, Red, agree with most of the conclusions of that book, right?

[teo123]

If you had no knowledge of Croatian whatsoever and blindly picked "river names" and only ever "river names" out of a hat with a million categories and found this association on your first try, that would be different.

OK, let's talk about things I came to know relevant to that k-r pattern after I've published the paper:
1. There are plenty of river names in Bosnia and Herzegovina that start with k-r. Multiple rivers named "Krupa"... I wasn't aware of that, but that does support my hypothesis. Illyrian was spoken in Bosnia and Herzegovina. I've tried to estimate the p-value, the upper bound (if you assume that the collision entropy of phonology is 3.195 bits per consonant pair) seems to be around 1/30, and the lower bound (if you assume the entropy of phonology is 1.623 bits per consonant pair) is so low my program ran into an infinite loop trying to calculate that.
2. There are a few large river names that start with k-r in Serbia. Karaš, Kereš, and Krivaja. Now, that's surprising to me. Illyrian wasn't spoken in Serbia. In Serbia in ancient times, Dacian and Thracian were spoken. And, unlike for Illyrian (scholars don't agree whether Illyrian was centum or satem), all serious scholars agree that both Dacian and Thracian were satem languages (the claim that Thracian was an Italic language is considered a fringe theory). If my theory that many river names come from *kjers is correct, we would expect, if anything, that river names on s-r are common in Serbia, rather than on k-r. But that's not what we see. That can be used as an argument against my theory. I can dismiss "Krivaja" as being probably Slavic in origin ("winding water"), but I cannot dismiss Karaš and Kereš the same way.
3. Yesterday, I came to know that "kurit" means "to flow" in the Silba dialect of Croatian. They have a saying "vrime kuri" (time flows). That supports my theory that *kurr was the Illyrian word for "to flow".

[teo123]

Other noun categories:
Town names
Hill names
Valley names
Stream names
Bridge names
People names
Tool names
Animal names
Business names
Whatever names

Can you find 17 general categories?

What do you mean "Business names" is a plausible category here? It's an extremely implausible category. Businesses in Croatia are mostly named after Latin and Greek words, so if a pattern such as the k-r pattern was found in business names, it would have to be a coincidence. The same goes for people names, which are mostly Latin, Greek and sometimes Hebrew in origin. Bridges are, as far as I know, mostly named after people, so bridge names are also a rather implausible category. Tool names... well, it's complicated. Some linguistic hypotheses are indeed based on patterns in tool names, such as the supposed Proto-Basque word *hai(n)tz for stone, from which tool names are supposed to be derived. Tool names in Croatian are mostly German or Turkish in origin, so a pattern in tool names would be rather difficult to explain if not as a coincidence. For river names (and probably mountain names), it seems at least vaguely plausible that a pattern such as the k-r pattern is not a coincidence (that there was a word such as *karr~kurr in Illyrian meaning "to flow" that many river names come from).

[brimstoneSalad]

What do you mean "Business names" is a plausible category here? It's an extremely implausible category. Businesses in Croatia are mostly named after Latin and Greek words, so if a pattern such as the k-r pattern was found in business names, it would have to be a coincidence.

You could control for those, or look at older ones.

The same goes for people names, which are mostly Latin, Greek and sometimes Hebrew in origin. Bridges are, as far as I know, mostly named after people, so bridge names are also a rather implausible category.

Remove those, and look at the others.

You also need to look at ALL possible patterns with respect to their probability, not just your particular observation. You're less likely to find any specific pattern than you are to find a random pattern that has the same p value as the one you have written on.

Given how many loan word and loan origins you're talking about here, you may have also just found an accidental pattern from another language that has been Croatian-ized.
Remember, rivers can often start or end in different linguistic regions, and those other language names may have spread thus bringing uncommon phonemes with them which influenced local naming convention and created a pattern that you wrongly ascribe a false cause of a different linguistic root to.

You need to find each of those rivers and follow them to see what they connect to; anything that flows into or (less likely) out of the region you have to count out of your sample because the names may be contaminated. You also need to look at how smaller rivers may have been misnamed for larger ones if the larger ones connect outside of the region.
See if the pattern still holds in the largest rivers originating and ending in the single linguistic region you're looking at (and which of course are traditional names which have not been changed recently).

[teo123]

You could control for those, or look at older ones.

So, do you think that Dubravka Ivšić is right to say that I would probably get a much larger p-value if I ignore the river names that have obvious Slavic etymologies (by which she means Krapina, Kravarščica and presumably Krbavica), and that that invalidates my conclusions?

[teo123]

@aroneous, you know something about the information theory, right? What do you think about my idea that information theory strongly suggests that this k-r that repeats itself in the Croatian river names (Krka, Korana, Krapina, Krbavica, Kravarščica, two rivers named Karašica) was the Illyrian word for "to flow"? I think the basic information theory suggests that the probability of such a pattern occurring by chance is somewhere between 1/300 and 1/17. I've published a paper about it in two "peer-reviewed" journals, and you can read about it on my website.

[teo123]

You need a P-value vastly disproportionate to the sample you drew from.
Not even 1/300 seems terribly convincing given that issue.

Don't you think the same response applies, if not much more so, to the studies supposedly showing that lockdowns work? There are different measures of lockdown strigency, some suggesting that Croatia had a more strict lockdown than Germany, the others (in my opinion, more realistic ones) suggesting the opposite. Furthermore, there are so many measures of success of lockdowns you can look at. You can look at how many people died with confirmed COVID-19, you can look at how many people the doctors think died from COVID-19, you can look at what's the excess mortality, you can look at the time of the peak of the curve... Then you can also control for the distance of the country from the equator, control for population density of the country, control for rates of type-2 diabetes within the country... Don't you think that, by chance alone, you will find a statistically significant correlation between some measure of lockdown strigency and some measure of the success if you search for it?

[brimstoneSalad]

You could control for those, or look at older ones.

So, do you think that Dubravka Ivšić is right to say that I would probably get a much larger p-value if I ignore the river names that have obvious Slavic etymologies (by which she means Krapina, Kravarščica and presumably Krbavica), and that that invalidates my conclusions?

Teo, I'm not an expert in linguistics nor even a novice, I'm only speaking generally from a perspective of informatics and statistics. I can't speak to specific words or assess the obviousness of anything with regards to its linguistic roots.

However, if you're asking if any randomly selected expert in linguistics is probably right with regards to a criticism of your work, the answer is yes, probably. Your work is probably wrong. I think even I have established that just based on your methodology without respect to specific analysis of your specific assessments (which I can not do).

Don't you think the same response applies, if not much more so, to the studies supposedly showing that lockdowns work?

I'm not going to go read those studies or humor conspiracy theories. That's too big of a rabbit hole. It's far more complicated than what you're doing, and I don't think an analogy is appropriate.

First, perfect lockdowns must work to eliminate a virus -- if you isolate all infected persons until they are no longer contagious, the virus is gone.
Whether lockdowns are successful in practice to slow viral outbreak (which was the point of them, to prevent hospitals from being overloaded) depends on how well they're followed by the people voluntarily locking down.
In practice it's probably nearly impossible to study the efficacy of various policies due to the confounding variables, but only nearly; you can get some sense of the probability of different policies having a better effect, and that can inform some controlled studies.
This is very different from what you're trying to look at, which is a post hoc association to find if there's a pattern at all rather than assessing a known mechanism to see if it's confounded in practice.

Stricter policies may be more effective, or less effective because people may be more likely to disregard them as too strict and impossible to follow so "why bother". It's a question of psychology and sociology. The data analysis on these is probably a nightmare. Then you need to compare them against other effects like depression and suicide that lockdowns could theoretically cause

Don't you think that, by chance alone, you will find a statistically significant correlation between some measure of lockdown strigency and some measure of the success if you search for it?

It's a bad comparison Teo.
But also no, not necessarily. The issue here in evaluating lockdowns is confounding human variables and adjustment for environment, not a random mistaken signal out of a huge pool of possible conjectures.

A better analogy would be something like looking at the success of lockdowns based on random variables, not known confounders -- like if you ad hoc hypothesized after looking at the data that lockdowns were most successful with people who have last names with even numbers of syllables including two ts, and then that was your conclusion. That would be stupid and nobody is doing that.

I don't want to discourage you from research Teo, but I'm getting the impression that despite your interest in linguistics, it may not be a good field for you.
You would be better off entering a field of harder science and limiting your research to actual experiments rather than ad hoc data analysis -- the latter of which will fuel your paranoid false pattern finding tendencies. I don't think this kind of linguistic research is compatible with or beneficial to your psychological issues. I would say talk to your therapist about it, but your therapist probably doesn't understand scientific methodology so probably would not be able to give sound advice on the topic. Just please consider shifting into hard science, you're still very early in your academic career and I don't think linguistics is healthy for you.

You could make some real contributions in the hard sciences and have a very valuable and successful career. I think if you continue in linguistics you'll be chasing phantoms and never really grasp the problem with it.