1 - the first step used by trouble-shooters is a straight checking of a closed system's actions. At its simplest it's asking "YES" or "NO" of all processes or decisions, with no exclusions.
John le Carré's `George Smiley' had the same method, using paper-records, to find the moles in MI-6* (S.I.S).
[Although geeks might call it `Boole's Logic', or "truth tables", when they're checking more complicated digital systems. http://en.wikipedia.org/wiki/Truth_table#Applications_of_truth_tables_in_digital_electronics ]
I.e. - basically it's collecting the data - on what the system does, and when.
2 - the next step is quantitative statistical analysis of the results you got from (1). This is the method used by insurance companies, political parties, gov't agencies & casinos - applications of it are often called `The Law of Large Numbers'.
`Law of Large Numbers' (Wiki)
Though not applicable to individuals, it's a powerful predictor and analyzer of the reactions of populations, and also of the declared outcomes of large organizations.
Warning: League tables of hospitals *, education systems *, police forces etc, claim to show the relative efficiencies of those systems. Mostly, that's not true - as seen below.
3 - the last step is qualitative or ontological analysis of the results you got from (2), by double-checking claimed or raw data outcomes against those predicted (demanded) by the `Law of Large Numbers' and by Bayes' Theorem.
Then there are a couple more analytical ways of identifying fraudulent social data (often `created' by corrupt local gov't officals, police, Edu and Health managers).
Because politicos' and bureaucrats' promises and their `official' outcome results often lie when viewed as `raw data'.
A legal example is at Bayes on drugs, a political example is at `damned lies' and an education example is at `examination bias'.
As that edu example shows, any mis-match is a sure indication of dishonesty and/or deception of some sort.
data - a system's actions or claimed actions;
indications - outcomes & claimed efficiency;
solutions - whether the system is honest or not.
As you might guess, politicians lawyers and judges (+ media) try to limit inquiry to step (1), maybe to hide inefficiency and hide corruption.
Examples - gov't criminals, fake fines, crapbbc, faked IFOs, sci-gov cover-ups
And some police/security, judges, scientists and bureaucrats try to restrict analysis to step (2), maybe to favor `selective' data, i.e. to hide bias and/or corruption.
Examples - `corrupt prosecutions' (text), `examination bias' , `violence check' and `pervs-in-gov't'
CAUTION re: `CATEGORIZING'
Categorization is a useful tool - when used after the analysis phase of a scientific investigation.
However, folk who've gotten opinionated (by the apparent efficiency of categorization) will often mistakenly categorize at the initial phase of data gathering. As a result they might declare `That's impossible / a hoax!' before analysis of data.
Such fatal `pre-judging' is a typical weakness of older, more dogmatic politicians, police, judges and scientists -
see Clarke's First Law
Here's some famous examples of eminent scientists making fools of themselves - by pre-judging the data, or categorizing before analyzing.
Pro-active Discrete Detection of Corruption:
is not for the squeamish, and can be physically dangerous.
The dangers, both physical and moral, of hands-on detection are obvious enough, because the investigator is necessarily consorting with those suspected of corruption.
Remote detection might seem safer but that isn't so. Corrupt agencies (including police) tend to react even more violently against `remote detection', because they fear the methods might become known to a wider public.
Methods of `remote detection'
1) Study and master all mechanisms involved - scientific or social or financial, and specially statistical, until qualified to hypothesize. [It's wise to conceal one's expertise and apppear `naive' until next stage is complete.]
2) Broadcast one or more speculative hypotheses, in purposely `naive' form - to rouse the corrupt agencies and, maybe more importantly, their hidden agents in society.
3) Note which speculation arouses most, and fiercest, opposition. [At this point one could reveal that expertise and expose those agents - or continue `naive questioning', broadcasting, and noting results.]