This is one of the worst papers I have ever read in terms of methodology. Honestly abhorrent.
1) Insanely low sample size.
1600 people to represent the whole segment of 18-35 is nuts. Any generalization is extremely risky, without even considering the rest of the paper, just on this notion alone. (I will explain why they ran this shit on 1600 people at the end of the comment, be prepared because it will make you cry due to the sheer amounts of cringe and disgust for the profession of "social sciences researcher").
2) Selection bias.
acknowledged by the researchers:
It definitely introduced selection bias. 100k registered members who get on those platforms solely for the incentives (money, gift cards) and realistically only care about optimization of revenue.
Justified as follows:
Detached from reality.
3) Social desirability bias.
acknowledged by the researchers:
Why? Because if you're asking people "hey have you committed this morally depraved and illegal act? we aren't collecting any info, source: trust me bro" you will always get downplayed results. Always. Both on the victim and perpetrator parts.
Not naming the provider of anonymous surveys = no guarantee that it was actually anonymous.
Here is the appended questionnaire:
The questionnaire has non-trivial issues:
1) Overlooking consumption of alcohol and drugs during parties. This is a major point that completely fucks up the statistical analysis, touched upon later;
2) Flattening a 1-10 spectrum of political opinions into a dichotomy (presumably 1-5 Left, 6-10 Right);
3) Flattening a 1-10 spectrum of such pornographic content as the above mentions, into a dichotomy (presumably 1 No, 2-10 Yes), which is is in extreme bad faith;
4) Not including female rapist's vagina/anus being forcefully penetrated with a male's penis (worded extremely poorly because I am retarded, it's the usual issue of rape laws being incomplete);
5) Not including certified diagnoses of psycho-pathologies such as ASPD, Psychopathy, Sociopathy etc. (which would be dubious unless rigidly certified through methods that would disrupt the anonymity of the respondents, rendering the study null), which have an empirical effect on the study's aim (as one can infer from the DSM-V-TR description of the symptoms of such psycho-pathologies, assuming the DSM is a good-willed manual and not (((their))) way of controlling people. I digress. This point also fucks up subsequent statistical analysis.
Technically called an "Omitted variables bias".
4) Cross-sectional study.
A cross-sectional study by its nature is unable to assess causal links.
Acknowledged by the researchers:
However, the researchers indulge in unwarranted hypothesizing that they themselves state being baseless:
5) The sample does not accurately represent the demographic.
The researchers seem to have forgotten to list the other factors that one can control in the "Results" section:
1) The real percentage of 18-35 year old Spanish with a uni degree is about 40%;
2) About 28% of 18-35 year old Spanish have a non-heterosexual orientation;
3) About 40% of 18-35 year old Spanish are on the political left.
4) About 65-70% of 18-35 year old Spanish have a low to medium or low socioeconomic status (€2000 monthly or less of net income)
5) About 20% of 18-35 year old Spanish are 1st or 2nd generation migrants from other countries
What does this mean? It means that the study is garbanzo-tier even before considering actual calculations. Why?
Researchers do not represent adequately the population of Spain aged between 18-35. Any statistical calculation done thenceforth is completely drugged by the skewed balance.
The initial sampling check is done, and it is not promising at all. Under these premises, the study should've been halted and remade from scratch with an actual representative sample and more precise representation via including factors such as certified ASPD, Psychopathy, Alcohol consumption during parties, Drug consumption during parties AT LEAST.
Oh but don't worry, for there is one particular passage that renders this whole paper scientifically dishonest, reveals the blatant HARKing and pushing of an ideological position and should make honest social scientists demand radiation for these "researchers". Of course, I will only reveal it at the end, for it will make much more sense then.
On to the actual data.
6) FUCK YOU MEAN 95% CI [1.72 - 8.28] ???????????????????????
I will assume throughout these sections that the reader does not know anything about statistics and linear regressions and allat bullshit (I barely know these things myself, but I know enough to talk about it).
Let's start from a quote:
I am falling into a state of madness.
We need to follow the process and then look at how they derived the results.
This is the key, this is where most of the shitfuckery goes on.
1)
Table 3. The relationship between lifetime drug-facilitated sexual assault (DFSA) perpetration while partying and the type of pornography consumed and frequency of consumption (N=1482). a
| Sociodemographic and behavioral variables | Perpetrated DFSA at any point in their life while partying | | P value | Crude ORb (95% CI) | Adjusted OR (95% CI) | | |
| Yes | No | | | | | |
| Type of pornography consumed, n (%) | | | <.001 | | | | |
| No consumption | 18 (3.6) | 485 (96.4) | | Reference | Reference | | |
| Without DFSA | 35 (4.6) | 719 (95.4) | | 1.31 (0.73-2.34) | 0.91 (0.43-1.95) | | |
| With DFSA | 47 (18.9) | 202 (81.1) | | 6.27 (3.55-11.06)c | 3.78 (1.72-8.28)d | | |
| Frequency of pornography consumption, n (%) | | | <.001 | | | | |
| Never to less than once per month | 33 (4.1) | 770 (95.9) | | Reference | Reference | | |
| Once per week to 2 to 3 times per month | 25 (7.5) | 310 (92.5) | | 1.88 (1.10-3.22)d | 1.01 (0.49-2.06) | | |
| Daily to 2 to 3 times per week | 42 (11.4) | 326 (88.6) | | 3.01 (1.87-4.83)c | 1.30 (0.65-2.59) | | |
| Sex, n (%) | | | <.001 | | | | |
| Female | 32 (4.1) | 757 (95.9) | | Reference | Reference | | |
| Male | 72 (9.2) | 711 (90.8) | | 2.40 (1.56-3.68)c | 2.03 (1.17-3.51)d | | |
| Age (y), mean (SD) | 27 (5) | 27 (5) | .79 | 0.99 (0.96-1.03) | 1.00 (0.98-1.05) | | |
| Sexual orientation, n (%) | | | <.001 | | | | |
| Heterosexual | 61 (5.0) | 1162 (95.0) | | Reference | Reference | | |
| Nonheterosexual | 42 (12.6) | 292 (87.4) | | 2.74 (1.81-4.14)c | 2.10 (1.33-3.32)d | | |
| Nationality, n (%) | | | <.001 | | | | |
| Spanish | 81 (5.7) | 1351 (94.3) | | Reference | Reference | | |
| Spanish and/or other | 20 (14.8) | 115 (85.2) | | 2.90 (1.72-4.90)c | 2.55 (1.40-4.62)d | | |
| Educational level, n (%) | | | .68 | | | | |
| University | 56 (6.9) | 759 (93.1) | | Reference | —e | | |
| Nonuniversity | 48 (6.4) | 707 (93.6) | | 0.92 (0.62-1.37) | — | | |
a The multivariate model included as covariates those variables that showed a
P value of <.05 in the bivariate analysis.
b OR: odds ratio.
c
P<.001.
d
P<.05.
e Variables not included in the multivariate model according to the results of the bivariate analysis.
THIS IS COMPLETE BULLSHIT
here's why.
The bivariate analysis works on a 2x2 Matrix of watching DFSA and perpetrating DFSA/watching DFSA and not perpetrating DFSA/not consuming and perpetrating DFSA/not consuming and not perpetrating DFSA.
The bivariate algorithm then calculates the odds of perpetratingg among consumers and divides it by the odds of perpetrating among non-consumers. With the data:
(47/202)/(18/485)=0.23267326732673267326732673267327/0.03711340206185567010309278350515=6.2692519251925192519251925192528 which is approximately 6.27.
HOWever. Take a glance at the 95% CI (Confidence Interval). This interval represents a very precise concept: 95 times out of 100 we sample the population with n=1482, the interval contains the true population parameter. What does this mean? It means that in reality, we do not know if this phenomenon is (under the bivariate, extremely imperfect algorithm for the study case) that there is a 3.55 factor increase in DFSAs among DFSA consumers, or a 11.06 factor increase. The superior extreme is more than 3 times the lower extreme, and even in social sciences, this is indicative that the model cannot predict the phenomenon. It would be like having a scale, and when we put a 627g object on it, the scale reads a value anywhere between 355g and 1106g. It has no predictive value. But the cOR of the bivariate is virtually useless here. What we need is a multivariate analysis because there are multiple variables at play here. Hence, the aOR.
The multivariate analysis is inherently flawed here because the major variables of "alcohol consumption", "drug consumption", "ASPD/Psychopathy/Sociopathy/other such psycho-pathologies" are ABSENT. This reflects on another value, the Pseudo-R-squared value, touched upon later.
The multivariate analysis is done, here, counting only the 5 factors that have a p-value less than 0.05 in the bivariate analysis. Hence, Age (.79) and Educational status (.68) are not inserted. We're left with a 5 variables function that iterates in a matrix of these values and maximizes the probability of observing the real distribution of the 1482 participants.
The algorithm calculates the beta coefficients of the variables and the aOR (adjusted Odds Ratio) is the exponential of the beta value (here, 3.78 for "Consumes DFSA porn".
Once again, consult the 95% CI: (1.72-8.28). There is no predictive capacity in this. It's either a mild increase, or a social plague catastrophe. The result is that "yeah, it's anywhere between 1.72 and 8.28" which is basically like saying "yeah there is anywhere between a 0 and a 100% chance of this happening" (exaggerated, but I need to get the point across of what this means in simple terms).
If you look at the row above, you see that the 95% CI has 1 in it. It means that there is absolutely no correlation (in this model) between watching porn of the non-DFSA kind and committing DFSA, something that was already visible in the bivariate.
Frequency of pornography is also irrelevant as pointed out by the researchers themselves.
What can also be seen is the stat on males, non-heterosexuals and immigrants.
So we have an utterly useless model for the purpose of the study (determining a correlative link between consumption of DFSA porn and perpetrating DFSA), with the results being "yeah anywhere between mild increase and full-blown apocalypse".
The p-value doesn't mean shit when it comes to these aORs and CIs, it merely means that "yes, actually, this model is this faulty".
And the ultimate proof is the McFadden Pseudo-R^2 being so abysmally low (0.123).
The McFadden Pseudo-R^2 is a type of check that is done to see if the model fits the data. Basically through it you can check the amount of variance, or "how much of the phenomenon is actually explained by the model". Here, a measly 12.3% of the phenomenon of DFSA is explained by the multivariate analysis, which just means this model is USELESS AS FUCK, because it leaves an 87.7% unaccounted (GEE I WONDER WHAT WOULD'VE HAPPENED HAD IT INTEGRATED OTHER VARIABLES HMMMMMMGE).
The chi^2 is useless as fuck here.
The first test is a Likelihood Ratio Test: "What's the probability that the likelihood of this multivariate (in this case, 8-variate) model is less than the likelihood of a 0-variate model (which is the null model)?". It being 88.8 with p<0.001 is due to the sample size of 1482, infinitely bigger than the 47 people that consume and perpetrate DFSA. This because chi^2 = 2*(ln(L_M)-ln(L_0)) where
L_M = likelihood of the model
L_0 = likelihood of the model without predictors (the variables)
now, ln(L) = sum from i=0 to i=N of ln(p_i) where N is the sample size, p_i is the probability of the event (in this case, perpetrating and consuming DFSA) calculated on each participant.
this means that the chi^2 value is directly proportional to the sample size. Its value of 88.8 is solely due to the sample size of 1482 obfuscating the irrelevancy of the model.
Why is the pseudo-R^2 this low, then? Because:
R^2 (of McFadden) = 1 - ln(L_M)/ln(L_0)
Applying the chi^2 formula here:
R^2 = -chi^2/2*ln(L_0)
What happens here is that even if the chi^2 value is apparently good, the denominator grows in a way that is directly proportional to N, the sample size.
Let's apply it to this study:
We need to find ln(L_0):
we know that ln(L_M)-ln(L_0)=88.8/2=44.4
we also know that 0.123 = 1 - ln(L_M)/ln(L_0) = [ln(L_0) - ln(L_M)]/ln(L_0) = -44.4/ln(L_0) so we find that ln(L_0) = -44.4/0.123 = -361 (approx)
which is coherent with the fact that L_0 is between 0 and 1 and logarithms of values between 0 and 1 are negative.
With this, we find that the L_M is -316.6 approx -317.
now, -chi^2 = R^2 * 2*ln(L_0) = 0.123 * -722 = -88.8 so that chi^2 = 88.8 as we have.
So?
So chi^2 is no parameter for judging the predictive ability of anything. It is merely a sign that "there is something in the data, and it's not due to white noise" the REAL parameter by which to judge predictive ability is the R^2, which is pathetic (0.123).
I need you to understand intuitively what is happening in this statistical hell.
There are 1482 people that constitute the sample. 100 of these perpetrate. 1382 of these don't. This is the base model without predictors, the null model. The equation that represents it would be log(p/1-p)=beta_0. beta_0 is a constant in the null model, also called "intercept".
The probability to extract a perpetrator out of the 1482 is 100/1482=0.067476;
The probability to extract a non-perp is 1382/1482=1-0.067476=0.932524.
Recall the earlier formula for ln(L). Here, there is a flat sum of 100*ln(0.067476) + 1382*ln(0.932524) = -366 (close enough to -361, approx errors and shit).
Then let's consider the 8-variate model. It assigns different probabilities to different cells in a matrix which represent intersectionality of variables. The likelihood is bound to be higher than that of the base model because of this, it is assigning "more correct" probabilities to different outcomes. The equation for this model would be log(p/1-p) = beta_0+beta_1X_1+...+beta_8X_8 (8 variables X_1 thru X_8 and 8 beta coefficients)
However, the summation is always done on the whole sample. This means that a low initial sample can have a chi^2 value below the minimum acceptance value (about 15.5 for 8-varied models) and signal that the model is shit as fuck, but a high initial sample with the same data of perps will pump up the model to high chi^2 values, however the R^2 will still be abysmally low to signal that the model is not explaining shit.
Recall the chi^2 formula: chi^2 = 2*(ln(L_M)-ln(L_0))
This is just two times the distance between ln(L_0) and ln(L_M), which is the difference in likelihood between the base and 8-variate model. What's happening? The 8vmodel (again, JFL if it wasn't) has a likelihood, ln(L_M), closer to 0 than the base model. This means the 8vmodel is getting rid of some chaos and providing explanation for some percentage of the phenomena.
The chi^2 formula has a minimum acceptance value, which for 8vmodels is about 15.5. What the chi^2 measures is simply if and by how much the proposed 8vmodel is "picking up explanations of phenomena".
The true measure of the predictive ability of the model is the R^2 though. By dividing the -chi^2 by the 2*ln(L_0) the R^2 is effectively measuring the percentage of "chaos" that is being "explained" by the model, essentially dividing the movement distance by the initial value. (In this case, a whopping 12.3% of it geg). It is always a number between 0 and 1 due to this.
The second test is the Goodness of Fit test of Hosmer-Lemeshow. It divides the sample in risk deciles and confronts the frequencies of "Yes" and "No" observed in the sample with those provided by the model. "Does my model fit the data well?". The higher the p value here, the better, because you're checking the truth value of the null Hypothesis (that the model fits the data perfectly) so it has to be closer to 1 than to 0. The problem is that with 47 people who consume and perpetrate DFSA, the sample is too scarce locally to be able to reasonably pick up the distortion of the model. Still, the p-value isn't even that high (p=.48 is basically a coin toss, which I wouldn't trust to fit the data personally).
This is the equation that represents it:
View attachment 1740692
- g ranging from 1 to 10 is the division in risk deciles, 1 being lowest risk and 10 being highest risk (as established by the 8-variate model);
- O_(1g) and E_(1g) are respectively the number of positive cases (in this case, perps) observed and expected by the 8-variate model.
- O_(0g) and E_(0g) are respectively the number of negative cases (in this case, those that do not perp) observed and expected by the 8-variant model.
What is happening here?
The 8-variate model assesses the risk for each decile. Then, it estimates (E_1 and E_0) the amount of people that will perp and not perp (1482/10=148 (rounded down) * risk probability calculated for that decile).
Then, for both the 1 (perp) event and 0 (non-perp) event, the distance between the observed and estimated values is squared and divided by the expected value, then they are summed and this for each decile.
What's the problem here? Without enough data, the lower risk deciles return irrelevant values that are in the order of 10^-1 or even 10^-2; let's assume the risk value of the 1st quintile to be 0.5%; then 148*0.005=0.74. That is the expected value for the 1 event. The expected value for the 0 event is 148-0.74= 147.26. Let us assume that the observable reality has O_(1,1)=0 and O_(0,1)=148.
The formula returns [(0-0.74)^2/0.74]+(148-147.26)^2/147.26=0.74+0.00371859296482412060301507537688=0.74 approx.
So when there is a lack of data (for instance, 100 perps in 1482 people), of course the chi^2_(HL) will be low (has to be lower than 15.5 for this test), returning a p value great enough to pass the test.
What would happen with more data?
Let's assume the sample size is N=12422 (
for no particular reason at all) with 385 perps (
for no particular reason at all)
Assume there were even just a handful of perps, say, O_(1,1)=3, say the risk value was 0.5% for the first decile: E_(1,1)=1242*0.005=6.21 [THIS IS MERELY TO SHOW HOW LACK OF DATA SKEWS THE TEST]
Do this shit for yourself, it's already fucking 4:40 AM and I've been writing this for 7 hrs, anyways the result is about 1.7 which is 2.5 times the original value. Very likely the chi^2_(HL) value will be higher than 15.5, completely fucking up the test. I'm tired. I wrote this after the below considerations and before the discussions section. the voices are winning ADHBAIDFHBCDFHBGA
Everything said above applies here too. With the only difference that this SHITTY fucking model accounts for... 6.6% of the variance in victimization. WHERE IS THE REMAINING 93.4%???? I already replied above as to why this is. Literally this model fails to explain why 93.4% of the DFSA happens irl.
This study is fucking bullshit. But before heading to the "Discussion" section, I want to highlight why this study should be studied in statistics courses as the most baseless, politically motivated, biased piece of shit ever seen ever.
50% PREVALENCE HAHAHAHAHAHAHAHAHAHAH
This was calculated using a formula that returns the minimum number of participants a study has to have to be valid, Cochran's Formula:
N=[Z^2*p*(1-p)]/E^2 where:
Z = 1.96 for 95% confidence;
p = Prevalence (here 0.5)
E = margin of error (here 0.025)
Plug the numbers in. The result? 1536.6 so 1537.
What's the problem here?
They used the prevalence value that would net them the least amount of participants to involve in the study (p*(1-p) has its maximum at p=0.5, assuming 0<p<1, extremes included)
However, assuming a prevalence of 50% for a literal fucking crime goes against basic principles of statistics regression. The standard is to assume a prevalence of 3% and a relative error margin of 10%
the corrected formula would be
N=[Z^2(1-p)]/((E^2)*p) where:
Z=1.96 for 95% confidence:
p= Prevalence (new: 0.03)
E = margin of RELATIVE error (new: 0.10)
The result? 12421.16 which is 12422.
TWELVE THOUSAND FOUR HUNDRER AND TWENTY FUCKING TWO PARTICIPANTS.
Why does this matter at all?
The Confidence Interval used in this study is calculated on Odds Ratios, so the extremes are not linear, but exponential.
the confidence interval is of the form [e^(beta-1.96*SE), e^(beta+1.96*SE] (1.96 is the Z parameter of 95% Confidence from before)
And the SE is approximated by the function 1/sqrt(x) (not in multivariates, the SE is higher due to the multivariate nature)
This means that the more data in the cell "Consumes DFSA porn and perpetrates DFSA", the less the value of the SE, and thus the less the amplitude of the interval, meaning the precision is higher.
These fucking retards, though, only had 47 data points. This effectively made the extremes of the CI explode and all the precision was lost.
With 12422 participants, though, assuming a found prevalence of 3.1% (WHICH IS CORROBORATED EVEN BY THIS SHIT FUCKING STUDY BECAUSE 47/1482 = 0.0317 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA THE VOICES THE VOICES AAAAAAAAAAAAAAAAAAAAAAA THEY'RE WINNING AAAAAAAAAAAAAAAA) there would have been 385 data points, and the SE would've been wayyyy lower.
In fact, let us calculate the SE with n=47.
SE = [ln(superior extreme of the CI)-ln(inferior extreme of the CI)]/(2*.1.96)=a bunch of numbers=0.4009 approx 0.4010.
This is coherent with a "Consumes DFSA porn" beta value that is ln(aOR)=ln(3.78)=1.32972400 approx 1.330.
Check = [e^(1.33-1.96*0.401),e^(1.33+1.96*0.401)]=[1.723, 8.2975] approx [1.72, 8.3] which IS EXACTLY WHAT THE RESEARCHERS FOUND.
As I said, this is incredibly irrelevant to any predictive capacity: "Hey I measured my temperature and the thermometer said my temp is anywhere between 35°C and 42°C!!!!". Worthless.
Let us assume the same conditions (3.1% prevalence, 95% Confidence, 10% Relative Error Margin, 3.78 aOR, 1.330 Beta Value, but n=385 perpetrators that consume DFSA porn (3.1% of 12422)) and let's run the calculations to see the new CI.
New SE = 0.401*sqrt(47/385)=0.1401.
New CI = MATH MATH MATH = [2.87 - 4.97]
way, WAYYYY more precise.
BUT.
This cannot realistically happen. This is assuming too many things:
1) "Linearly-Geometrically" Diminishing SE value. In reality, with a sample nearly 8 times larger, the covariance would likely completely alter the values, the aOR would not still be 3.87, the beta value would change, and thus the CI.
2) Introducing the "forgotten variables" (mentioned like 5-6 times already) would introduce causes that are empirically much, much more relevant than these ones (as one could assume by the 0.123 McFadden value JFL, 87.7% of the cases of DFSA cannot be explained by this model JFL JFL JFL)
3) Having the sample reflect accurately real life demographics would likely increase other covariances' relevancy, further diminishing the aOR of "Consumes DFSA", and I think [PERSONAL OPINION] the CI's inferior extreme could even be veeeeeeeeery close to 1, from above.
So, what do we actually know?
1) This model is worthless.
2) Any conclusions one can draw from this model are inappropriate as a result.
3) Psychiatry is not a science and social sciences are not scientific at all.
4) Torture data, and they will tell you any story you want to know, or something.
Discussion Section
Holy JFL.
proved to be faulty as fuck and likely irrelevant when compared to the "forgotten variables"
so? this was not the point of the study at all.
DFSA pornography isn't necessarily more violent than any other type. I don't want to dismantle each of the studies, let's suppose they are reliable. Then what? How does this data tie into your study? You have "proven" that there is no precision in the data and no conclusion can be drawn off of your faulty fucking model. Good job. Again, you cannot explain 87.7% of the DFSA perpetrators' variance and 93.4% of the victims' variance. Your model is incredibly useless. YOU YOURSELVES have established the statistical irrelevancy of the Confidence Interval of frequency and type of porn watched, barring DFSA and DFSA ONLY, YOU DID NOT EVEN ACCOUNT FOR "VIOLENT PORN".
Your aOR is faulty, your R^2 value is enough for me to discount your shit.
How are these scripts perceived as normative or acceptable and how do they explain the 87.7% and 93.4% that you couldn't account for?
KEK FUCKING MAO at the politicized double standards.
"Men? Duh, of course they are here".
"Faggots, troons and niggers? WOAH WOAH WOAH, there are 69 googolplexes relevant factors like COUNTRY OF ORIGIN and the HETEROGENEOUS NATURE of those groups (because everyone knows men are one entity silly, take your lithium it's time for your (((meds)))). Let's not get ahead of ourselves here. 10 million more euros in bullshit research to produce worthless papers are needed."
ARE YOU FUCKING SERIOUS RIGHT NOW??????? YOU'RE JUST GOING TO GLOSS OVER ALCOHOL AND DRUGS????????????????????????????????????????????? WHEN YOU CAN'T EXPLAIN 93.4% OF THE VICTIMIZATIONS??????????????????????????????????????????????????????????????????????????????????? CAN YOU FUCKING STOP LOOKING AWAY FROM REALITY FOR A SECOND???
WHAT???? YOU'RE ADMITTING THE ILLICIT NATURE OF DRAWING CAUSAL HYPOTHESES, YET DOING SO ANYWAYS??????????????????????? AHAHDUSHDQJKWAD BGEQKFHG BWEIFKRJFHGBSRUIKFH BRHJKFHGDJHFHJDHGFJDHAHHAHAHHAH DUOHDIUEHFWEFH THE VOICES HAHAHAAHAHHAHAHAHHAAHAHAHAHAHAH
DUDE, YOUR FUCKING WORTHLESS MODEL literally shat out a p value greater than .05 for educational level. AND COUNTRY OF ORIGIN? YOU DIDN'T ACCOUNT FOR THAT DJQW DHBQLKIU DHGDHHDHDHDOIHD.
Do you understand this quote?
They're saying they don't give a fuck if they did a cross-study, that cannot reveal causal links. They don't give a fuck if pornography consumption is not even correlated with perpetrating DFSA. They don't give a fuck. (((They))) will "implement the necessary sexual health promotion interventions" because that was the entire goal of this bullshit paper. Provide an excuse, even a poorly crafted one, for policy. Don't believe me? Look at who funded this paper. I will tell you in a minute.
AAAAAAAAAAAAAAAA POLITICS POLITCSIS QASAAASPAJWDIòOJQAWDIOJHQWAIKDUHJDIQAUWKHDIQKAHD AAAAAAAAAAAAAAAAAA (thanks for the funding daddy UN and EU) AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA.
provides literally no evidence.
no it DOESN'T.
I'll leave the commenting to you. It's 5 AM. 8 hrs writing this shit. wtv. more online legislation. more censorship. more power to (((them))) and women. women most affected btw.
should've made it public for ease of access and trust.
postdoc grants are fucking cancerous, NextGenerationEU is post-covid funds (5 yrs later still pumping funds OK), HOW IS THIS HELPING TO RECOVER FROM COVID ASDHJNQWOAHD NQLIAHD ILUHIUDF. Also legit funding from the "University Institute for Research in Police Sciences", but absolutely no involvement in the study (trust me bro) GEG.
I want to die so fucking bad ChatGPT and OpenAI written papers SJWIOPDJEOòFCHJWLIFUKWHFUWEHF.
Why do I think they HARKed the study?
It's not really HARK (Hypothesizing After Results are Known) but more of a conditioned study (hehe I fucking baited you HAHAHAHAHAHA):
Nobody had thought of studying this shit before; there was no internationally validated questionnaire for DFSA perpetration and porn consumption; the statistics are extremely jarring and misleading and the model is worthless; the ideological tendencies are very potent; funded by the literal "University Institute for Research in Police Sciences"; pushing for legislation and "sexual health awareness" stuff when you have a literal nothingburger; avoiding alcohol and drug consumption as covariates; shit sample that doesn't represent the Spanish population, instead represents liberal leftards university partygoers, literal fratfagboys; hastily crafted sample size likely due to high cost of finding 12422 people that could represent the spanish population decently and maintaining a study with them, applying a textbook technicality that shouldn't apply when investigating crimes; will to infer causal links from a cross-sectional study; prolly more shit I'm forgetting.
if you want, read this thread I made:
Fuck introductions, fuck segues, fuck everything you niggERs are going to say. Straight to the point. 1) Free will is a delusion that the System exploits to better control the herd aka normie soyciety. Logically speaking, the matter of free will and determinism is strictly undecidable from...
incels.is
if you want, read why I think psychiatry is retarded and the DSM is stupid cultish propaganda:
Nowadays it's almost ubiquitous seeing people blame men for women's problem or call them misogynist if they dare criticize a woman. Whilst feminism is seen as a legitimate academic subject, anything related to men's rights just gets labeled misogynist and associated with the alt-right. Barely...
incels.is
Nowadays it's almost ubiquitous seeing people blame men for women's problem or call them misogynist if they dare criticize a woman. Whilst feminism is seen as a legitimate academic subject, anything related to men's rights just gets labeled misogynist and associated with the alt-right. Barely...
incels.is
I am not a social scientist, just a mathematics undergrad. Feel free to dismantle my statistics knowledge if you know more than me. Tell me directly if that is the case. this study made my eyes bleed.
. The difference between Spanish and foreigners/mixed ancestry(?) people being even bigger than between the sexes is also interesting, and implies that minority and foreign women are doing so more than Spanish guys.
