6.5
Risultati insoliti
I risultati insoliti sono quelli che hanno una probabilità molto bassa di verificarsi. I risultati insoliti possono essere identificati utilizzando le…
I risultati insoliti sono eventi che hanno probabilità molto basse di accadirsi. Possono essere identificati con la regola empirica dell'intervallo o con i valori di probabilità.
Si consideri la distribuzione di probabilità dell'occupazione dei posti in un carpooling con una media di 3,5 e una deviazione standard di 1,2.
Secondo la regola empirica dell'intervallo, la maggior parte dei valori delle variabili casuali deve trovarsi entro due deviazioni standard della media.
Tutti i valori di dati rimanenti che non rientrano in questo intervallo sono valori insoliti.
Per identificare risultati insoliti dai valori di probabilità, si consideri la distribuzione di probabilità del numero di teste in una moneta lanciata cinque volte.
Poiché la probabilità di zero o meno teste è inferiore a 0,05, questi risultati possono essere etichettati come insoliti.
Allo stesso modo, se la probabilità di cinque o più teste è inferiore a 0,05, anche questi risultati sono insoliti.
Il valore di cut-off di due sigma per le variabili casuali e 0,05 per la probabilità non è rigido. Può essere deciso in base al contesto del problema.
View the full transcript and gain access to JoVE Core videos
Q1: What makes a result unusual in a probability distribution?
Unusual results are events with very low chances of occurrence. They fall outside the typical range of values in a probability distribution. Using the range rule of thumb, values beyond two standard deviations from the mean are considered unusual. Alternatively, results with probabilities less than 0.05 are labeled unusual, though this threshold can be adjusted based on context.
Q2: How do you apply the range rule of thumb to identify unusual values?
The range rule of thumb states that most random variable values lie within two standard deviations of the mean. Values falling outside this range are unusual. For a carpool with mean occupancy of 3.5 and standard deviation of 1.2, the unusual range extends beyond 0.1 to 6.9 seats. Any observed value outside these bounds indicates an unusual result.
Q3: What probability threshold typically indicates an unusual result?
A probability value less than 0.05 typically signals an unusual result. For example, in a coin tossed five times, getting zero heads or five heads each has probability less than 0.05, making these outcomes unusual. However, this 0.05 cutoff is not rigid and can be adjusted depending on the specific problem context and research requirements.
Q4: How do probability values and the range rule of thumb differ in identifying unusual results?
The range rule of thumb uses standard deviations to identify unusual values: results beyond two sigmas from the mean are flagged. Probability-based identification directly examines outcome probabilities, labeling those below 0.05 as unusual. Both methods identify low-likelihood events but use different mathematical approaches. The choice depends on whether you have a probability distribution or raw data.
Q5: Can the cutoff criteria for unusual results be modified?
Yes, the cutoff criteria are flexible and context-dependent. The standard two-sigma threshold and 0.05 probability level are conventional but not rigid rules. Researchers can adjust these thresholds based on problem requirements, field standards, or acceptable risk levels. Different contexts may warrant stricter or more lenient criteria for classifying results as unusual.
Q6: What are the two types of unusual outcomes in probability problems?
Unusual outcomes occur as either unusually high or unusually low numbers of successes. In a binomial probability distribution, both extremes—very few or very many successes—represent rare events. These opposite tails of the distribution contain results with low probabilities, making them statistically unusual regardless of direction.
Q7: How do you calculate the unusual value boundaries using standard deviation?
The unusual value boundaries are calculated using the formulas: maximum unusual value equals mean plus two standard deviations (μ + 2σ), and minimum unusual value equals mean minus two standard deviations (μ - 2σ). Any observed value exceeding these boundaries is considered unusual. This method provides a clear numerical range for identifying outliers in random variables.