I am a statistician and software engineer, living in Israel.

My interests include politics (theory and practice) and statistics (theory and practice).

Within politics, I am focused on issues revolving around elitism, political equality, and the application of chance in order to achieve political equality. I am an editor and a contributor to a group blog, Equality-by-Lot, which deals with the use of chance in governance and resource distribution.

If you have a statistical problem *associated with a pro bono cause*, please write to me at.

Advertisements

January 19, 2009 at 7:26 pm

[…] בטיפול בבעיה : פרסום מידע המדויק כפי שנתן לי בנדיבות יורם גת : התשובה הקצרה לשאלה שלך היא שהתפלגות ההכנסה בישראל […]

May 27, 2009 at 7:54 pm

Assuming you are “Sortition”

Thanks for your reply on Yglesias page regarding your solution to the Israel-Palestine problem, especially Jerusalem. I apologize for not responding earlier, I was away.

It may, or may not, surprise you to know I hold the same view as you. I think there might need to be some compromise solution regarding the Jewish quarter and the Western wall in the old city.

My whining on about complexity in the situation has less to do with the notion that a solution must be inherently complex than the popular notion (often promoted by the media on the left and right) that Israelis all have one view, and Palestinians all have another.

Finally, I spent some time in Abu Ghosh, an Arab village in Israel, and found it both uplifting and depressing (!) It seems to be the place that works. Jews shop there, eat at the restaurants there, sit and chat there. The Arabs work and eat and shop on the Jewish side of town. It’s the town that (mostly) works. Uplifiting for the obvious reason. Depressing because the solution is staring everyone in the face. Naive, I know, but I just can’t get up every morning and be cynical.

(Interestingly, if the country returns to ’67 borders, some residents on the Jewish side of town will have to move…)

Thanks again for your thoughtful reply.

May 29, 2009 at 6:04 pm

Hi Mark,

Yes – I was writing as “Sortition”. Thanks for your reply and for dropping by.

I am not surprised that our positions on the settlement of the conflict are similar – both being close to “the international consensus”. This is the settlement that any reasonable person would support under the current conditions, and it is even given lipservice to by Israel and the US. Any disagreement between us is probably regarding the question of who is responsibile for the current impasse. US and Israeli media and governments place the blame on the Palestinians and many people tend to assume that there is at least some truth to that. In fact, a look at the fact reveals that the impasse is wholly created by Israeli intransigence, which is made possible by the backing of the US.

Regarding Jerusalem: the Palestinians have already made in 2000 at Camp David the concession of offering control over the Jewish Quarter and the Western Wall to Israel as part of the peace settlement. (Of course, the Israelis demanded more.)

Regarding Abu Ghosh as a model for co-existence: I don’t quite agree. There is a conflict of interests here. “Why don’t we just all live peacefully together?” pretends that this is all just a silly misunderstanding. This is not the case. The Arabs in Abu Ghosh and elsewhere in Israel are an oppressed minority, not equal citizens. Israel has been created by subordinating the interests of the Palestinians to those of the Jews – necessarily so. Co-existence involves a compromise – it is not only about being open-minded.

July 22, 2009 at 8:21 pm

Just wanted to say I just found your blog today quite by accident searching for statistics on the US income distribution and it’s very interesting! thanks!

April 3, 2013 at 2:54 pm

Hi Mr Yoram Gat,

Your professional interests remind SoundEagle of computational governance and probabilistic politics as well as stochastic resource distribution.

January 24, 2014 at 5:12 am

Hullo Yoram,

To start with, thankyou for your altruism herein.

I thought I might bounce this particular thesis off you for your critical review.

As you are aware, we conduct citizen juries for government and we are starting to use a sample size of 43, selected at random (from the community in question).

We have found that these juries regularly obtain at least 80% consensus on the matters that are being deliberated: so we are using a Confidence Interval of 15%.

Here below is a short precis from another author on this methodology and I was wondering whether you might agree or otherwise.

Thankyou.

Luca

PS 43 is very close to the answer to everything in “The Hitchhiker’s Guide to the Galaxy”…that was 42!

“Firstly, a representative sample is randomly assigned from the community population to participate in the jury. In choosing the sample and estimating the appropriate effect size, the statistical concepts of confidence level and confidence interval (margin of error) are important considerations (Cumming, 2012).

A confidence level is a percentage that conveys the level of certainty that the estimate is equal to the value that would be obtained from measuring the total population. In the social sciences, the confidence level of 95% is routinely used, and for this particular study is deemed appropriate, i.e., there is a 95% level of confidence that any decisions the jury make are representative of the total population (Cummings).

The Confidence Interval (CI) is effectively the range within which results can be expected to fall, if the measurement is repeated. The larger the CI range, the wider the margin of error. The APA (American Psychological Association, 2013) states that researchers should routinely report CIs and also use them as the basis for interpreting study results. In

Secondly, the CI can be used to determine the sample size for the study using a sample size calculator readily available online (The National Statistical Service, nd.). If this sample is too large to recruit and work with in the community engagement process, using a 15% CI brings the appropriate sample size down to only 43 people. However in this example it is critical to ensure that at least 80% of the sample agrees with the proposition; the wider margin for error (with the greater CI) allowing for the possibility that as few as 65% of the population agree with the proposition or as many as 95% agree (80% +/- 15; Ellis, 2010).”

References

American Psychological Association. (2013). Publication manual of the American Psychological Association (6th ed.). Washington, DC: Author.

Cumming, G. (2012). Understanding The New Statistics: Effect Sizes, Confidence Intervals, and Meta-Analysis. New York: Routledge.

Ellis, P. D. (2010). The essential guide to effect sizes. Statisitical power, metaanlysis and the interpretation of research results. New York: Cambridge University Press.

The National Statistical Service. (nd.). Sample size calculator. Retrieved 12 January, 2013, from http://www.nss.gov.au/nss/home.nsf/pages/Sample+size+calculator

January 25, 2014 at 8:15 am

Hi Luca,

I think the treatment dives into the mathematics too quickly, glossing over fundamental matters have to be carefully considered first.

Confidence intervals are used in the context of estimating the proportion in the population of a certain characteristic. For example, what proportion of the population supports a certain candidate. In the current context, it is far from clear what the characteristic whose proportion in the population we are estimating is. Unlike a regular opinion poll, where the possible answers are given by the pollster, we are considering a situation where the sampled people are generating the answers. This makes the standard analysis inappropriate, at least a-priori.

Another issue that is ignored here is that requiring super-majorities for adoption of new policy in effect privileges the status-quo, which may be quite unpopular. For example, if the adoption of new policy requires support by 80% of the sample, and the existing policy enjoys the support of 25% of population, then those 25% will like have a representation in the sample that can block the adoption of an alternative policy that is supported by the remaining 75% of the population.

January 30, 2014 at 1:00 am

Hi Yoram,

Thanks for this feedback.

I agree that one needs to make the minipublic/ descriptive representation (and cognitive diversity) arguments first. We’ll do that,

But we also need to make a statistical claim, otherwise interest groups will just play an arms race of numbers – ie MoveOn have 500,000 supporters and we only have 43, so therefore MoveOn are always right.

On the link between confidence intervals and a single characteristic: what we are promoting is that the characteristic in question is judgment (or “where do the jury find the reasonable tradeoff”) .

So, if we repeated the exercise with another sample but with the same information then there is the likelihood we would get the same type of answer/judgement.

Our experience is that juries rarely go to a vote, (if there are dissenting voices, they’re less than 10% of the jurors) So the issue of >20% minority positions rarely arises. We’re just using the 80% metric for the statistical argument.

Does this all make sense?

Luca

January 31, 2014 at 6:17 pm

Hi Luca,

> So, if we repeated the exercise with another sample but with the same information then there is the likelihood we would get the same type of answer/judgement.

Unfortunately, it is difficult to evaluate success at the test you propose. Since the wording of two policy proposals cannot conceivably be identical, and since given two non-identical proposals it is impossible to decide objectively whether they represent the same “answer” or “judgment”, there is no objective way to know whether the test was met successfully or not.

I suggest a different test, then:

given a proposal and the process which produced it, what proportion of the citizens, being well-informed about the proposal and the process, would consider them appropriate, i.e., representing the public interest?This is a test that most present-day elected parliaments fail horribly at. Large majorities of the population in Western countries distrust their parliaments and consider both the policy they generate and the way they generate it as being inappropriate. Yet, as I see things, a decision making process and the decisions it produces can be considered democratic only if most people, after becoming informed about them, find them appropriate. We need to design a decision making process that does meet with (informed) public approval.

So, here is my practical suggestion. You should split your sample into two parts: (a) the policy-making sample, which would discuss and vote and finally propose policy, and (b) the monitoring sample, which would monitor the policy making sample and finally decide whether its process and proposal represent the public interests. Now you can have the rigorous statistically valid argument that you seek. If the monitoring sample has a statistically significant majority pronouncing that the policy-making sample has done its job well, then you can expect a majority of the population to feel the same (under the same circumstances). Note that the formal statistical calculation (the confidence interval) is carried out on the monitoring sample, rather than on the policy-making sample.

In fact, you could also have a monitoring sample which observes an elected body and produces an assessment of how well the elected body represents the public interest. The informed assessment of the work of the elected body could be of great interest in itself and could also be interesting as a point of comparison for the assessment of the work of policy-making sample.

If you are interested in putting something along those lines into practice, I’d be happy to help.

April 1, 2017 at 3:49 pm

[…] Gat, one of our new Contributing Editors, describes himself in his blog (also called Pro Bono Statistics) like this: “I am a statistician and software engineer, living in Israel. My interests include […]