Musings tagged as statistics

Trump Beaten by Facts (Again)

Fri, Jun 22, 2018

Trump tweeted that crime in Germany is “way up” and that this is due to all the refugees taken in by my country. While I would quickly run out of hard drive space if I were to comment on every single time that dimwit claims something despite the evidence pointing in the completely other direction, this one - of course - is a bit personal.

Luckily, we Germans love our statistics and cold hard facts. For anyone interested you can check out the official crime statistics of the Federal Statistical Office of Germany. As these only go up to 2016, here are the latest numbers for 2017, you can find these on page 10 in the PDF file, section 3.1. The total number of criminal offenses registered in 2017 is 5,761,984. This is the lowest number since 1992 and a good 10% less than in 2016.

This was way too easy… But then again, if you are Trump you already are your worst enemy anyway. The rest of the world can just sit back, relax, and enjoy some popcorn while the US fade from the world stage.

I’d never thought I would say this but Uber, the ride hailing company, did something remarkably awesome. A group of five economists, two of which employed by Uber, two Stanford professors, and the chairman of the University of Chicago economics department have released a paper in which they report on their analysis of more than 740 million Uber trips in the States between Jan 2015 and Mar 2017, involving more than 1.8 million drivers.

The price a customer has to pay for an Uber ride is calculated by an algorithm that does not care about gender. The deciding parameters in making up the price of the fare are trip distance, wait time, speed, and surrounding circumstances like scarcity of available drivers. Even though there is no gender involved and the algorithm computing the fare is not just completely neutral in that regard but also does not care about things like whether someone works part- or full-time, funny enough there is still a gender pay gap. According to the paper, men earn an average of $21.28 per hour while women only earn an average of $20.04 per hour. The difference of $1.24 amounts to a gender pay gap of about 6%.

How can this be? There is no evil patriarchic society at play, the math is simple (the equation is actually part of the paper) and it does not discriminate against anyone. In their analysis, the writers name three main causes for the paygap that can be proven scientifically by the ride data they analyzed.

1. Men have more experience

The authors state that men and women learn at the same rate in terms of number of rides. They also state that for example wait times go down by 5% to 10% over 1500 rides of experience for both genders. This is because both men and women learn about which rides to reject and which to accept. But, according to the statistics, men learn more intensively per week of experience as they work longer hours. After a certain time interval, men will have accumulated more rides than women and thus more experience. With passing time, the percentage of men with a lot of experience will rise faster than the percentage of women. Also after six months, 77% of women will have quit working for Uber. With men, only 65% will have quit in the same period, leading to a further increase in high experience male drivers.

2. Men drive faster

For both genders, the speed goes down with experience as the drivers learn that congested areas are more lucrative than being out and about in the countryside. But, men still drive faster on average than women. They also drive longer trips and the combination of longer trips completed in less amount of time means more money. The authors mention studies that show that men are more risk tolerant than women, both in general and when driving in particular. This might explain the general tendency to drive faster. While mostly irrelevant in the daily life, in a drivers line of work speed of course pays off.

3. Men pick better spots and ride times

Possibly also a matter of experience, men tend to favour areas that have a lack of available drivers even though there is high demand. This leads to a bonus modifier for the fare in order to get more drivers into areas where they are needed the most. Men more actively seek out areas where there are high bonus factors available, leading to more income per trip.

What do we learn from this?

One of the most deciding factors is time spent working. If women prefer to work part-time they will accumulate less experience. Less experience usually means being less productive which then results in less pay. Either directly if the wage is coupled to hours worked or also indirectly because someone with more experience will move up the food chain faster. This holds true for both genders and this report shows that women do not need men to discriminate against them in order to be paid less. They can achieve this just fine by themselves. The question one should ask would be why the difference in work hours? Family? Lack of interest? Maybe being an Uber driver is not appealing to the general female populace, hell, I could hardly think of a more annoying job myself. It’s all in the eye of the beholder I guess.

But, if someone suggests that a person A with less experience than person B should be paid the exact same amount, they are actually the ones who are discriminating.

And since we men are apparently born with the need for speed and a greater risk tolerance, take it easy girls, this also means that we are far more likely to live life the squirrel way: Live fast, die young, and leave a flat patch of fur on the highway ;)

Every time another study claims to have found a link between an increased cancer risk and mobile phone usage, like the newest NIH study here (TR-596) I cringe in my chair, awaiting the hordes of inept journalists to jump on the bandwagon and proclaim practically the end of civilization due to death by smartphone. Of course if the media is writing about it then the average Joe is inclined to believe it, even though this question can actually be approached with some common sense.

So what is cellphone radiation? Cellphones emit microwaves, which is the designation for electromagnetic radiation with a wavelength between 1 mm and 1 m or a frequency between 300 MHz (3 * 108 Hz) and 300 GHz (3 * 1011 Hz). Which frequency is emitted by your cellphone depends on the network and technology in use. Most phones operate around 900 MHz, 1900 MHz, or 2100 MHz, depending on the frequencies in use in your country. Here is a nice graphic that shows the range of the electromagnetic spectrum by the way. So, how can this harm us? If you are wondering whether the term microwave refers to the same device that you use to heat up your food then you are correct, but hold your horses for a minute, it all depends on the power! And before we come to that, let us examine what ways there are to affect tissue in a way that this could lead to something like cancer.

Destroy the DNA with ionizing radiation

A surefire way to get cancer is to break up the bonds that hold your DNA together. While your body has ways to repair broken DNA, it sometimes makes mistakes while doing so. The more mistakes happen, the higher the chance that the DNA will be distorted in a way that using its (now errorneous) information to produce otherwise useful things for your body, now leads to the formation of a malignant tumor that grows and spreads. So the first and most obvious way to get cancer is to cause as many breaks in the DNA as possible.

Your DNA is held together by ionic bonds which have a certain force of attraction if you will. If something highly energetic comes along, think of something like a bullet, and hits this bond, it can break. However, your bullet needs to have a certain amount of energy so it can actually overcome the ionic bond or nothing will hapen. Think of taking a cannonball and throwing it against a house wall with your hands. You can probably do this all day and nothing much will happen. Load that cannonball into a cannon first and fire it at the house and it will go right through. This is of course because the cannonball has more (kinetic) energy when fired from a cannon since your arms are just too weak to give it sufficient energy.

It really is the same with electromagnetic radiation or photons, which is one and the same. A photon hitting your DNA must have a certain amount of energy or nothing will happen. Throwing more than one low-energy photon at your DNA does not help either in the same way as it does not help if you and your friends throw cannonballs at a wall together. More balls do not make up for one ball with enough energy to penetrate…. let’s use a different analogy next time. :o) Long story short, more particles do not help, they need to have enough energy in the first place!

Can mobile phones destroy your DNA?

So, how much energy do you need to destroy DNA? The range of ionizing radiation starts at around 120 eV which is short for electronvolt and is the unit physicists like to describe the energy of particles with. There is no need to understand the background of electronvolts or how to convert it into different units of energy as you will soon see, it is all about the ratio. So everything above 120 eV will be able to destroy your DNA but the more your are below 120 eV the less likely this can happen.

The higher the frequency of a photon, the higher its energy so lets take the highest frequency we have for mobile phones which, as I wrote above, is 2100 MHz. Let’s make it 3000 MHz just for kicks. At that frequency a photon will have the energy of 12 µeV which is microelectronvolt. Micro is a millionth so a mobile phone microwave photon has 12 millionth of 1 eV of energy. How many eV did we need to destroy our DNA? Around 120. This means that photons emitted from mobile phones only have one ten millionth (0.0000001) the amount of energy that would be needed to ionize our DNA bonds and lead to cancer this way. This will never ever happen! CASE CLOSED

What about the heat?

Another way to get cancer is by heating up your tissue in excess of 42 °C which leads to proteins to denature, a fancy way to describe killing cells if you like. This is also one of the reasons why a really high fever can be life-threatening. So if you kill cells, your body has to replace them. When building new cells, your body can fuck up, so to speak, and put something together the wrong way. If the resulting misbuilt cell starts to multiply like crazy and spreads around, there’s your cancer again. So, since we cannot ionize the DNA with microwave radiation, can we maybe heat up the tissue enough for cells to die so they have to be replaced? After all, your microwave oven can also heat up your food pretty good.

This is where the amount of photons comes in. While more photons hitting your tissue do not raise the chance to destroy DNA if they do not have enough energy to do so, they do warm up your tissue. This amount of photons is what we call power and we usually use watts (W) to describe it. Your microwave oven usually has something in the order of 800 W of power. This means a whole lot of photons hitting your food, which then gets really hot. You really should not stick your head in there which is why microwave ovens have a safety switch so they can not be turned on if the door is still open. (Put down that screwdriver and leave that safety switch alone! It’s there for a reason!) Your mobile phone has a lot less power!

The current SAR (specific absorption rate) limit for exposure to radiation is around 1.6 W/kg or 2.0 W/kg depending on the limits enforced in your country. This basically describes how much power per kg of tissue is deemed safe. The SAR rating of your phone will be even less. Most phones are in the range of 0.2 W/kg to 1.0 W/kg. If the reception is good the phones also throttle down to save battery, their power output in that case is even only a fraction of these values. There is a nice website from the Bundesamt für Strahlenschutz (the German agency for protection from radiation) where you can look up the SAR values of hundreds of phones (available in English). (Please don’t mind them using the German word for mobile phone which is Handy. So if they ask you for your handy model don’t get any funny ideas, just input the model of your cellphone ^^)

As we can see, mobile phones only have tiny amounts of power compared to a microwave oven (Lucky us!). Is this enough to maybe at least get your tissue to 42 °C? There has been research conducted by Van Leeuwen, et al. who developed a head temperature model based on MRI scans to investigate the effect of microwave radiation on the tissue temperature. On top of it all, they also verified their model by experiments they conducted to measure the skin temperature rise, so this is a pretty solid piece of work.

The blood circulating in our veins is the main source of cooling and heat transport in our body, else we would overheat. Also, simple physics tells us that the more difference in temperature there is between two areas or volumes, the more the heat transport is increased. This also means that as tissue warms up, the temperature difference to the blood temperature becomes larger and thus the heat transfer to the blood gets stronger resulting in even better cooling. If you have a constant heat source, the temperature rises at first but then, as the cooling becomes more efficient, levels out and stays at a constant value without increasing further unless you apply more power. Van Leeuwen found that the maximum rise in brain temperature after hours of constant mobile phone usage was 0.12 °C. A bloody tenth of degree! The average body temperature of a human can easily vary by as much as 1 °C throughout the day. How is this supposed to create cancer problems for you? You should be more affraid of getting cancer from catching a cold with the resulting fever.

So have fun, go about your business, use your phone as much as you want, and stop worrying. Also stop fucking telling people that they get cancer. Oh, before I close this, how did the newest study I mentioned at the very top determine that there is an increased risk for cancer? Well first of all they blasted mice with 10 W/kg of radiation. Your body cooling system is efficient but it will crap out at some point. You can not simulate a long exposure by increasing the dosage if a lower dosage is completely safe in the long run. We have a built-in “air conditioning” and of course you can make it fail if you make it hot enough. They could’ve put the mice into a microwave oven and then complained about health side-effects, same deal.

Oh and they don’t seem to be so sure of themselves either:

The combined incidences of fibrosarcoma, sarcoma, or malignant fibrous histiocytoma of the skin were increased in 5 and 10 W/kg males, although not significantly or in an exposure concentration-related manner.

Great, not only did we up the dosage to an unrealistic level, it also hardly shows any difference…

James Damore Sues Google

Tue, Jan 9, 2018

James Damore, the engineer that got fired by Google for writing an internal memo about how there is an aura of fear amongst Google employes of being fired for expressing conservative or even just scientific views about the whole gender debate (note the irony), has just filed a case against his former employer (and here is the suit) for discrimination.

It will be interesting to see how this pans out. Apparently the suit (linked above) contains numerous screenshots of interesting things going on. In addition to marvels in the text, as in Google paying a so called peer bonus where a colleague can recommend another colleague for a bonus because that colleague spoke out against the values in Damore’s memo. Oddly reminiscent of Trumps let’s-buy-UN-votes tactics.

Overall, the second half of the suit is riddled with screenshots of things posted on the internal google newsgroups and message groups. It actually paints a pretty clear picture of who is discriminating against who.

A team of security researchers has developed an algorithm with which they can fool neural network based image classifiers, such as Google’s Cloud Vision, with a remarkable success rate of > 95 %. They can actively change the classification result by generating an image which looks like A but gets classified as B. In their publication it is shown how they successfully fooled several neural network image classifiers to think that a picture showing a couple assault rifles actually shows a helicopter. A picture of a guy on a snowboard and a guy on skis is classified as a dog.

In a nutshell they take a reverse approach and start with an image that shows the adversarial (fake) object, for example a helicopter. This image of course gets classified as such. Then this image is modified over several iterations to look different, for example like a couple assault rifles, while still retaining its classification as a helicopter. In the end they show that they can pretty much make anything be labeled as something completely different with a success rate of more than 95 %. The combinations are staggeringly confusing:

  • a cat gets labeled as an airplane
  • an airplane gets labeled as a deer
  • a deer gets labeled as a truck
  • a lionfish gets labeled as eggnog (this could really hurt ^^)

One thing these images have in common is that they sort of lose their (fake) touch if they get transformed, for example rotated by more than 30 degrees. Also we are talking images so who cares about someone being able to fool a neural network into thinking a perfectly aligned 2D image of a rifle is a helicopter?

Well, they did not stop there as their newest publication deals with robust adversarial examples. In reality this culminates in them printing a 3D model of a turtle which is classified as a rifle regardless of background or rotational angle. Oh, and they made a baseball look to a neural network like an espresso. Check this screenshot of page 8 of the most recent paper:

adversarial thumb

Essentially they went the same route as before, modifying the texture to look like A but be classified as B. Albeit with more tweaks so it keeps up the sharade even when rotated.

I think this is remarkable, especially in light of neural network based image classifiers being used more and more not just in everyday life but also for security purposes. So if you can reliably make a security system think that an assault rifle is actually a helicopter, a teddy bear, or a cup of espresso, just by painting or printing a certain texture on it, you pretty much won the game there.

This also tells us it still pays off having a human brain, for now at least.

Apparently there is another new buzzword (at least for me) you can write down for later use in your bullshit bingo sessions. Gender Pricing is the name of the game and I have just been alerted to this by reading about a recent gender study (German) by the Antidiskriminierungsstelle (federal anti-discrimination agency) that attempts to find cases of women being charged more for the same product or service than men.

Their posterchild is the visit at the hairdresser where women are usually charged more than men. Now this is probably already obvious to you if you have ever been to the hairdresser with your wife or girlfriend: A womans haircut can hardly be compared to a mens haircut unless the woman gets a quick neckshave with an electric razor and a little bit of scissor action on the top. Notice I am not even mentioning coloration, streaks, bleaching or whatever here (which takes the whole case even more ad absurdum) just the cut in general. This is also the main reasoning given by the hairdressers that were interviewed in the framework of this study as written in the aforementioned report on page 122:

Even though the argument of the businesses, that women haircuts require a significantly larger amount of work, can be confirmed by observations and questioning of female customers, there are only limited choices for women. For example a woman, who only wants a simple “mens haircut” will usually not dare to ask for or receive it.

So even though all evidence points to a significantly higher workload for womens haircuts than for mens, and this is confirmed in this study, the very few women that only want a quick electric razor shave do not get one and one of the reasons is that they do not ask for it. And this is supposed to be the fault of who exactly? In addition, even though I am sure they exist, I do not know a single woman who would be satisfied with one of these super simple mens haircuts. So now the hairdressers are sexist because they do not immediately cater to the insignificant number of razor-cut loving females that are too scared to ask for one themselves? But wait, there is more:

Questioned about the reasons for the increased workload when catering to womens hairdressing needs and regarding the amount of time scheduled for these appointments the businesses replied that the service for women is more time intensive. Reasons for this are many, for example the complexity of the haircut, the need for more consulting and discussion, the lower frequency of visits and therefore more work to be done at each visit, and that haircuts regarded as “short” by women are still significantly longer than those regarded as short by men. If the time spent on customers of different sexes is taken into account, the difference in price is negligible. Usually 30 minutes are reserved for men and 45 minutes for women. Even with all these reasons one still has to question the reasoning behind these differences in time and especially because of the difference in service the pricing could be done without taking the sex into account.

So after being given loads of valid reasons, which were confirmed by the female customers before, explaining in detail why a womans haircut is more time consuming, which then completely kills the difference in price, one should still argue against women being alotted more time than men. So apparently the hairdresser should just work faster when serving women? At least that would be the only way I can think of that would solve this weird request. And they finish by stating that especially because of all the different (and expensive) things you can pay for at your hairdresser the pricing should be the same for men and for women.

What kind of an idiotic request is this? I am pretty sure that if a man with a meter of hair books an appointment at a hairdresser in order to get streaks, some bleach, and a delicate haircut culminating in a complicated updo after that will not get this service for the same 20 EUR that some other guy paid for just getting a 10 minute razor haircut.

This study is actually contradicting itself on pretty much every page unless it finds that “there isn’t any observable difference after all”. And if the hairdressers story wasn’t embarrassing enough, on page 102 you can find pictures and prices of comparable sets of razor blades for mens and womens razors with a headline that reads

Example for non-equally priced personal hygiene products: razor products. Prices for the Aldi-razors (left) 4.49 EUR “for women”, 3.89 EUR “for men”.

The only problem is, the prices shown on the product pictures at the bottom of the page are 1.43 EUR for women, 1.59 EUR for men. I don’t know if you have to be a mathematician or physicist to recognize which sex pays less here for razor blades. Granted, this is the “discounted price” but even if you look at the non-discounted price it would be 4.79 EUR (women) and 4.99 EUR (men) which still favours the product marketed to women over the one marketed to men. In addition these are completely different numbers than the claimed 4.49 EUR and 3.89 EUR in the text. Who the hell proofread this pamphlet? (Here’s a screenshot in case they move the PDF or change it)

Of course the German media is all over this and I wish journalists would at least try to read and check a study before they take it for granted. The persistence with which the authors try to make a case against all facts, and the bluntness they employ while doing so up to openly admitting they have no case but wanting to complain anyway, is almost frightening.

What really amazes me is that with every “terrorist attack”, even when essentially not much happened in terms of casualties (no deaths, 4 lightly injured plus a moderately injured “terrorist”), people and media (german) are sent into a doomsday frenzy of “The terrorists! They are going to kill us all!”. The usual suspects chime in about how immigration is the root of the problem and that we need more laws, more surveillance and then there is this pesky internet:

Anyone can go on the internet and download garbage and vileness on how to put together an amateur-level explosive device – Andrew Cuomo, Gov. of NY

Since I like logic, reasoning, and most of all facts, let’s have a look at how threatened we are by terrorism in the western world, in comparison to other dangers around us. According to the National Vital Statistics Report of the CDC, in 2014 (latest data with this level of detail) we can look up the mortality rates for that year in a lot of different categories. Let us not count suicides and natural causes and then just pick a few not too exotic ones.

cause of death count
cancer 591,700
diabetes 76,488
influenza/pneumonia 55,227
drugs 49,714
traffic 37,267
falling 32,024
alcohol 30,722
firearms 12,208
suffocation 7,239
malnutrition 3,933
peptic ulcer 3,037
drowning 3,034
cutting/piercing 1,860
pregnancy/childbirth 1,123
Number of deaths by cause in 2014 in the United States

Looking at this data, if I were an american citizen, I would most of all be concerned with death by cancer, food or sugar, getting a cold, drugs, or drinking. Now if you excercise regularly, eat a good diet and have a healthy immune system, don’t do drugs and don’t drink, you should at the very least be affraid of driving, walking or otherwise taking part in traffic. Not to mention falling on your face and of course being shot by your fellow countrymen, …‘MERICA! (sorry, couldn’t resist ^^). More than 3000 deaths by malnutrition is also something that I think is crazy in todays time in a western country but back to the topic at hand.

How many people would you guess died due to terrorism in the United States in 2014? Look at the table for a second and try to guess where the terrorism row is supposed to go. According to a report by Miller and Jensen from the START consortium in 2014 the number of deaths due to terrorism amounted to 19. Yes, that’s right, nineteen! If you count the last 10 years in the report we end up at 186. By the way, this includes all deaths attributed to terrorism in the United States in general. That means this not just includes the islamic terror but also deaths due to other terror “organisations” like white supremacist groups, etc.

“But Alex!”, I hear you say, “What about 9/11?” And of course you are correct. If we count all deaths due to terrorism from 1995 to 2016, which is the whole range of the START report and which naturally includes 9/11, we end up at a gross total of 3393 deaths, 3003 of which due to 9/11, within 21 years. On a yearly basis that would be 162 deaths per year putting it at the very bottom of the cause of death table. You are 11 times more likely to die by cutting yourself than to die in a terror attack. If you are female, you are almost 7 times more likely to die during pregnancy or childbirth.

Out of the 2.6 million people dying each year in the United States about 200,000 people die due to accidents and violence, that is non-natural causes excluding illnesses like cancer. Out of these, 162 a year are killed due to terrorism, which makes for a terrorism death ratio of 0.081 %.

How do things look in my own country of Germany? Our media is babbling on about the danger of terror attacks and our politicians are calling for more surveilance and mandatory backdoors in encryption software. Surely the number-crazy Germans have their statistics right and don’t just give in to the spread of hysteria? Well, according to the GENESIS database and the Cause of Death ICD-10 statistics of the Federal Statistical Office (Statistisches Bundesamt) Germany we can get the data for the same categories as for the US. In addition, since both data are categorized according to the international ICD-10 standard released by the World Health Organization (WHO) they are absolutely comparable.

cause of death count
cancer 226,337
diabetes 24,401
influenza/pneumonia 20,874
alcohol 13,889
falling 12,867
traffic 3,688
peptic ulcer 2,436
drugs 1,558
suffocation 1,400
drowning 444
cutting/piercing 193
firearms 112
malnutrition 86
pregnancy/childbirth 24
Number of deaths by cause in 2015 in Germany

And how many deaths related to terrorism in the same 21 year period? Exactly 51 according to the same source (START) as above. Granted, more than half of that (27) happened in 2016 but in this year, which is almost over, we had 1 (one!) death due to a terrorist attack. Same as in 2015 by the way. Still, lets count everything since 1995 as we did above and we end up at almost 2 deaths per year!

In Germany, around 900,000 people die each year of which about 36,000 do so due to accidents, injuries, and violence. The same non-natural causes as I mentioned above. Out of these 2 people die due to terrorist attacks. This results in a terrorism death ratio of 0.006 %!

To put both countries into perspective, while the risk of dying due to a terror attack is 15 times higher if you live in the United States instead of Germany, you are also 20 times more likely to be shot by your neighbour. ;-) Germans however are 2.5 times more likely to suffer death by alcohol but then again, drinking beer is basically required by law over here!

Now, things are for sure a lot different if you do not live in the western world but instead in a country where the numbers in the tables are a bit different. If I were to live in Afghanistan I would definately be concerned about terrorism but the only cure that really helps against pretty much everything is education! Educated people are the antidote to extremism, regardless of which kind, and power hungry politicians alike. And while they are at it, they can also fix all of the things above that are way more likely to harm you.

As a closing thought, since, according to the data above, you are a lot more at risk of dying due to a stomach ulcer than being blown to smitherenes by some guy with a half-assed pipebomb down his pants, there is one thing I can recommend which is known to help: less stress, more chill, take it easy, and don’t get too excited about imaginary dangers! If you want to be concerned, be concerned about what your politicians do under the guise of the threat of terror.

And don’t go about fixing my beer… I like it just the way it is.