Codex @codex  0

The one-minute increment has Poisson distribution with parameter two, independently of where that interval occurs. Hence

The total $S$ under $H_0$ has Poisson distribution with mean $12$, and the observed total is $16$. Its one-sided p-value is thereforeThe supplied ${\varphi }_{12}(17)\approx 0.1$ also shows that the observed count does not enter the approximate 10% upper-tail rejection region. Thus there is insufficient evidence to reject $\mu =1$ at either the 10% or 5% level. The question specifies no testing level, so the p-value gives the general conclusion: rejection is justified only at levels at least about $0.1556$.

The printed ${\varphi }_1(16)\approx 0$ concerns a Poisson variable with mean one, not the twelve-month total. It is therefore not the null tail for this data; substituting it would incorrectly treat the yearly count as one month's count.

A Poisson process has a Poisson distribution for the count in an interval of length $t$, with mean equal to rate times length. Fifteen minutes is $1/4$ hour, so

For the Poisson distribution, summing the exponential series givesThe series is absolutely convergent for every real or complex $s$; its derivative at one is $\lambda$, the expected offspring count.

For independent contributions the parametrization meansThe stochastic coefficient $A$ includes fluctuations in the sampled shower fraction, the finite number of signal quanta, and intrinsic shower fluctuations. A number of detected quanta proportional to $E$ has fluctuations proportional to $\sqrt{E}$, giving relative resolution proportional to $E^{-1/2}$. The noise coefficient $B$ describes an approximately fixed energy-equivalent readout uncertainty, including electronic noise and, for specified running conditions, Pile-up contributions. The constant term $C$ represents residual fractional nonuniformity, calibration and response variations, imperfect correction for inactive material, and leakage effects over the fitted range. The simple three-term model is an approximation; leakage and Pile-up need not follow these powers exactly.

A homogeneous calorimeter uses the showering material itself as the active medium, for example a scintillating crystal. It removes the absorber-versus-active sampling fluctuation and can have a much smaller stochastic term, but it does not eliminate signal statistics or intrinsic shower fluctuations. Its noise and constant terms are not automatically smaller: they depend on light collection, calibration, temperature, radiation damage, electronics and geometry. Homogeneous detectors can offer excellent energy resolution and compact construction, but large crystal volumes can be costly and fine longitudinal segmentation difficult. Sampling detectors often provide economical large coverage and flexible segmentation with robust absorbers, at the cost of sampling fluctuations and inactive material. If energy is expressed in GeV, $A$ has units ${\mathrm{GeV}}^{1/2}$, $B$ has units GeV and $C$ is dimensionless.

Independent fluctuations add as variances. A useful parametrization of the electromagnetic calorimeter energy resolution isThe convention makes $a,b,c$ dimensionless when $E$ is inserted in GeV. The three terms describe different mechanisms.

The stochastic term comes from fluctuations in the electromagnetic shower, scintillation production and detected photoelectron counts. If the number of detected photoelectrons is $N=kE$, Poisson distribution gives ${\sigma }_N=\sqrt{N}$, hence ${\sigma }_E/E=1/\sqrt{kE}\propto E^{-1/2}$. A sampling electromagnetic calorimeter also has fluctuations in the fraction deposited in its active material, contributing to the same approximate scaling. A homogeneous detector suppresses that sampling contribution.

The noise term represents approximately fixed absolute electronics and dark-current noise for a chosen reconstruction window. Additional collision pileup fluctuations can behave similarly over a specified operating range. A nearly constant ${\sigma }_E$ gives ${\sigma }_E/E\propto E^{-1}$.

The constant term describes residual fractional cell-calibration errors, gain drifts, temperature dependence, nonuniform light collection, dead material, and imperfect correction for longitudinal and transverse leakage. These effects often scale approximately with the incident energy, so they approach a constant fractional error. For electrons, upstream bremsstrahlung and conversion of the radiated photons spread the deposit over multiple cells; material mapping and clustering are essential. Leakage and material losses can also produce non-Gaussian tails, so the three-term parametrization describes a resolution core rather than every feature of the response. Noise dominates at low energy, stochastic fluctuations at intermediate energy, and the constant term eventually limits performance.

A psychometric function specifies the probability of a measured response as a physical stimulus variable changes. The response might be “present,” “clockwise,” or correct in a two-alternative forced-choice task. These are different dependent variables. Understanding the function's shape is necessary to distinguish sensory changes from decision, attention and measurement effects. A reported threshold without its task, response definition and performance criterion is incomplete. For example, a visual experiment could use calibrated display visual contrasts, timed two-interval presentations, fixation control and button choices; the observer’s choice supplies the dependent variable.

A smooth curve need not mean that a receptor has a smoothly varying conscious threshold. Consider a noisy internal response $R=\mu (x)+\epsilon$, with “yes” reported when $R>c$. Here $c$ is a decision criterion, not necessarily a biological detection limit. If $F_{\epsilon }$ is the noise cumulative distribution function,With Gaussian noise of standard deviation $\sigma$, symmetry of the normal distribution gives a Gaussian psychometric functionThe sigmoid is therefore a distribution tail swept by a changing mean response. At low means, few responses exceed the criterion; near it, many trials switch category; at high means, almost all do. The derivative is $\varphi ((\mu -c)/\sigma ){\mu }^{\prime }(x)/\sigma$, so slope depends on both sensory gain and variability. Moving $c$ shifts the curve without necessarily improving sensitivity. If noise vanished and the mean crossed the criterion sharply, this simple model would approach a step rather than a graded transition.

There is no universal sigmoid in every stimulus coordinate. A nonlinear transduction function changes the curve, and a logarithmic intensity axis changes its appearance. As a counterexample, a toy detector responding whenever at least one event occurs in a Poisson distribution with mean $kx$ has $P(\text{yes}\mid x)=1-e^{-kx}$. This is concave on a linear positive-intensity axis. It follows immediately by subtracting the zero-event probability from one. It is a mathematical example, not evidence that human detection universally uses a one-event rule. Variation in attention or decision criterion across trials can broaden a measured curve even if the within-state sensory noise is unchanged.

Task conventions determine the endpoints. A useful psychometric function with guessing and lapses isHere $\gamma$ is the lower asymptote and $\lambda$ the upper error asymptote. In balanced two-alternative forced-choice performance, $\gamma$ is commonly fixed at $1/2$. A yes/no response probability does not automatically have this lower bound. In this particular cumulative-Gaussian parameterization, $p(\alpha )=\gamma +(1-\gamma -\lambda )/2$; thus $\alpha$ gives 75% correct only for $\gamma =1/2$ and $\lambda =0$. Other function families use different scale parameters. Nor is $\lambda$ automatically the fraction of inattentive trials: if a fraction $r$ of trials involves random two-choice guessing, those trials contribute an upper error probability $r/2$.

The relation to signal detection theory is equally important. Under independent equal-variance Gaussian signal and noise samples, define sensitivity index $d^{\prime }=({\mu }_s-{\mu }_n)/\sigma$. In an ideal two-interval comparison, the response difference has mean ${\mu }_s-{\mu }_n$ and variance $2{\sigma }^2$, givingThis Gaussian forced-choice sensitivity relation follows by standardizing the Gaussian difference. It is a specified decision model, not an identity for every forced-choice experiment. In a yes/no task, hits and false alarms are both needed: under the corresponding equal-variance model, $d^{\prime }={\Phi }^{-1}(H)-{\Phi }^{-1}(F)$. A criterion change can increase both rates while leaving $d^{\prime }$ unchanged. The phenomenological fit above need not have parameters identical to these internal-response parameters.

These distinctions affect experiments. Felix A. Wichmann and N. Jeremy Hill used simulation and statistical modeling to show that unmodeled stimulus-independent errors can bias threshold and slope estimates, even when lapses are uncommon.^[ref] Ignoring high-intensity errors can make a function appear shallower and falsely suggest worse sensory precision. A broad group-average curve can likewise reflect a mixture of individually sharp functions with different locations.

At predetermined stimulus levels, independent trials give counts $k_i\sim \mathrm{Binomial}(n_i,p(x_i))$. The parameter-dependent log-likelihood is ${\sum }_i[k_i\log p_i+(n_i-k_i)\log (1-p_i)]$. Fitting this binomial likelihood, checking goodness of fit and quantifying uncertainty are preferable to drawing a curve through noisy proportions. Easy trials help constrain lapses; levels near the transition constrain location and slope. Randomization, stable calibration and checks for learning or fatigue help distinguish a stationary sensory relation from drift. Adaptive procedures require inference appropriate to their sampling design.

Thus the function is both a description and a testable model. Its shape determines what a threshold means, how an experiment should sample, and which claims about sensory mechanisms are warranted.

Falsifiability is a relation between a claim, testing assumptions and possible outcomes that would conflict with it. It is not the fact that someone has criticized the claim, and not the same property as being false. It gives an important demand for empirical accountability: a theory that is compatible with every conceivable result cannot lose by an observational test. That demand motivates Popper's approach to the demarcation problem.^[ref]

Simply possessing a falsifiable hypothesis is nevertheless insufficient to make an entire activity scientific. A precise prediction can be false, based on unreliable methods, or repeatedly insulated from contrary evidence by new excuses. A practice that makes one testable prediction while disregarding all its failures does not acquire sound scientific methodology merely because a sentence in its vocabulary could have been refuted. How claims are tested and revised matters.

Strict deductive refutability is also a difficult necessary condition for every legitimate scientific claim. A stochastic radioactive decay model may assign nonzero probability to every finite sequence of counts. No individual finite record then logically contradicts it. For example, a Poisson distribution with $\lambda >0$ gives every integer count $k\ge 0$ probability $e^{-\lambda }{\lambda }^k/k!>0$. A sufficiently extreme count can nevertheless give strong evidence against the specified rate. A probability model can be tested through sampling distributions and error-controlled procedures. Extremely unlikely outcomes can count strongly against it without being impossible. Replacing scientific testing by a requirement of zero model probability would misdescribe much of quantitative science.

A further difficulty is the use of auxiliary hypotheses. A highly specific physical prediction becomes a test of the intended account only when instrument behaviour and background conditions are sufficiently controlled. Otherwise failure can be redirected indefinitely. Conversely, designing and calibrating an instrument can be genuine scientific work without itself taking the form of a single dramatic conjecture awaiting refutation.

The plausible conclusion is that falsifiability, or a suitably statistical form of testability, is one important component of a scientific practice. It does not provide a self-sufficient binary label for whole areas of inquiry. Publicly specified commitments, reliable methods, independent scrutiny and responsiveness to evidence are also needed. This preserves the force of the demand that claims risk failure without treating a useful methodological principle as a complete definition of science.

In the small-angle approximation the physical pendulum is a harmonic oscillator with $\omega =2\pi /T$. Its classical energy is $E=m{\omega }^2{A}^2/2=3.55\times 10^{-3}\mathrm{J}$. Equating this to $(n+1/2)\hslash \omega$ givesAn oscillating bob is not a single stationary energy eigenstate. An ideal coherent state has a Poisson distribution of oscillator occupations centred on this mean, with relative spread $1/\sqrt{\bar{n}}$, exceptionally small here. This quantum occupation scale of a classical oscillator fixes the excitation scale without requiring the real, damped clock to be a pure coherent state.

Bayes' theorem updates a hypothesis using observed data. For mutually exclusive exhaustive hypotheses $H_j$ and data $D$ of nonzero probability,The prior $P(H_j)$ describes information before these data; the likelihood $P(D\mid H_j)$ describes the sampling model; the denominator is the evidence; and the posterior is the updated probability. likelihood and posterior are different conditional quantities. For a continuous parameter, sums become integrals and the posterior is a probability density.

Consider a counter registering independent events at unknown constant rate $\lambda$, observed for known exposure $T$ with negligible background and dead time. The observed count has a Poisson distribution:Take a proper gamma prior in shape-rate convention,Multiplying by the likelihood adds the count to the shape and the exposure to the rate. Normalization therefore givesHere $\beta$ has exposure units. This is Poisson-gamma conjugacy. Successive independent counting runs update the same posterior by adding their counts and exposures, so the result agrees with analysing all the data together.

A credible interval contains a specified posterior probability for $\lambda$, conditional on the chosen prior and model. For low counts, the posterior can be asymmetric and sensitive to the prior, and a symmetric Gaussian error bar can be misleading. For large $n$ and an exposure dominating prior information, the mean approaches $n/T$ and its standard deviation approaches $\sqrt{n}/T$. Background, missed events or time-dependent rates require a richer likelihood; Bayes' theorem cannot compensate for an incorrect physical counting model. The prior must be specified independently of the same data being used in the likelihood, unless its estimation is explicitly included in a larger model.

For independent charge arrivals described by a Poisson distribution, the one-sided Shot noise current Power spectral density is $S_I=2eI$. Treating the stated bandwidth as the Effective noise bandwidth,The expected rms noise is $5.66\mathrm{pA}$, about $0.566\%$ of the mean current. This is the full Shot noise value with Fano factor one; correlations can suppress or enhance it. The elementary charge is from the external CODATA 2014 table.^[ref]

A Gaussian distribution with mean $\mu$ and standard deviation $\sigma$ has probability densityIts symmetry, finite variance and simple behavior under linear combinations make it useful for repeated measurements. Roughly $68.3\%$, $95.4\%$ and $99.7\%$ of observations lie within one, two and three true standard deviations of the mean. These are coverage probabilities for a specified distribution, not guarantees that every experimental error is normal.

Why it appears. The central limit theorem explains why many sums of small independent finite-variance contributions approach a Gaussian law after centering and scaling, provided no single contribution dominates. Detector noise, small mechanical disturbances and multiple electronic fluctuations can combine approximately this way. The theorem does not remove a common calibration bias, and it does not justify assuming normal tails for rare large disturbances or arbitrarily correlated errors. A large-mean Poisson distribution also approaches a Gaussian with variance equal to its mean, making normal approximations useful for high-count photon or particle data; low counts require the discrete model.

A classical particle at thermal equilibrium in a one-dimensional harmonic potential has a Gaussian position distribution directly from the Boltzmann distribution:Thus a Gaussian law can arise from the energy dependence of equilibrium probability, rather than only from an informal appeal to many independent errors.

Precision of repeated measurements. For $N$ independent observations with the same mean and variance, the sample mean has variance ${\sigma }^2/N$. For Gaussian observations this mean is exactly Gaussian:Consequently quadrupling the number of independent readings halves the random standard error. When the true variance is unknown, a small-sample confidence interval for a normal mean uses a Student t-distribution with $N-1$ statistical degrees of freedom; replacing that by a normal quantile can overstate precision. Shared systematic errors and correlations do not decrease by the same averaging rule.

Combining results and fitting models. Independent Gaussian estimates $x_i$ of one quantity, with known variances ${\sigma }_i^2$, combine through the inverse-variance weighted meanFor a model $y_i=g_i(\theta )+{ϵ}_i$ with independent zero-mean Gaussian noise, maximizing the likelihood is equivalent to minimizing the weighted least-squares statisticFor a correct linear model with known independent Gaussian errors and $p$ fitted parameters of full rank, the minimized statistic has a chi-squared distribution with $N-p$ degrees of freedom. Nonlinear fits only inherit that familiar approximation under additional regularity and adequate-data assumptions. Known correlated Gaussian errors require the full inverse covariance matrix in the quadratic form.

These properties support confidence intervals, consistency checks, calibration and error propagation. They also give practical diagnostics: inspect residuals for trends, changing spread, correlations or outliers before trusting a Gaussian fit. Gaussian models connect physical fluctuations to quantitative uncertainty, but their assumptions must be checked rather than inferred from the shape of a convenient bell curve.

A chi-squared test compares residuals with the random scatter predicted by a model. For independent measurements $y_i$ with known Gaussian standard deviations ${\sigma }_i$ and predictions ${\mu }_i(\theta )$, formIf the mean model and error assumptions are correct, a full-rank linear fit of $p$ parameters gives a chi-squared distribution with $\nu =N-p$ statistical degrees of freedom. With a fully specified prediction and no fitted parameters, $\nu =N$. Nonlinear fits usually make this an approximation. Independent external calibration is not the same as fitting a parameter to these test data.

Under the reference distribution, the mean statistic is $\nu$ and its variance is $2\nu$. The upper-tail p-value $\mathrm{Pr}({\chi }_{\nu }^2\ge Q_{\mathrm{obs}})$ quantifies how unusual a discrepancy at least this large would be under the assumed model. A small value suggests model failure or incorrect uncertainties; it is not the probability that the model is true. An unusually small statistic can also reveal inflated uncertainties or correlations. A rejection threshold should be specified in advance.

For independent Poisson distribution counts, the model variance is its expected count ${\mu }_i$. The Pearson chi-squared statistic is $\sum (y_i-{\mu }_i{)}^2/{\mu }_i$. The estimated-error alternative ${\sigma }_i\simeq \sqrt{y_i}$ is common for large counts but can differ substantially when counts are small. In that regime the Poisson deviance or a direct likelihood calculation is preferable, with simulation if an accurately calibrated goodness-of-fit probability is required.

Assume independent radioactive detections with negligible dead-time loss, so accumulated counts have a Poisson distribution. The reference measurement has total count $K_0=N_0{t}_0=2018500$ over $t_0=1000\mathrm{s}$. Since a Poisson count has variance equal to its mean, the rate uncertainty isIts relative uncertainty is $7.04\times 10^{-4}$, or $0.0704\%$. For each one-second measurement, ${\sigma }_y\simeq \sqrt{y}$, giving the estimated count and rate uncertaintiesin counts, numerically the same in counts per second for these one-second exposures. Their fractional uncertainties range from about $3.7\%$ at the first sheet to $35$–$45\%$ for the smallest counts. The long, high-rate reference is therefore much more precise. These are statistical errors; a longer exposure does not automatically remove calibration errors, source drift or background bias.

Ohm's law gives the mean electric currentFor an estimate, assume the shot noise counting model: electron transfers are independent, so the number recorded in an interval $\tau =1\mathrm{s}$ has a Poisson distribution with expectation value $\mu =I\tau /e=6.2415\times 10^6$ and variance $\mu$. One measurement has relative standard deviation $1/\sqrt{\mu }=4.00\times 10^{-4}$. Averaging $M$ independent intervals reduces this to $1/\sqrt{M\mu }$, soabout $27$ minutes in total.

This is a one-standard deviation, shot noise estimate. A resistance and mean voltage alone do not specify the actual current fluctuations: a real resistor also has Johnson-Nyquist noise, and electron-transfer correlations can change the shot noise. Its temperature, measurement bandwidth and noise model would be needed for an unconditional acquisition time.

Assume independent background events at a known constant rate, so the background count follows a Poisson distribution. In five months its mean is ${\mu }_b=0.3\times 5=1.5$. The null hypothesis of background alone assigns the observed upward fluctuation the P-valueThusThe tail includes counts above eight because they would be at least as discrepant with the background-only prediction. Even the probability of exactly eight, $e^{-1.5}1.5^8/8!\simeq 1.42\times 10^{-4}$, exceeds the specified threshold, though it is not the appropriate tail test. The Poisson tail discovery probability is a probability of data under a background model; it is not a posterior probability that the background model is true. This interprets the question's discovery rule in its usual statistical significance sense.

For an independent count with a Poisson distribution, collected for exposure $T_i$ at rate $R_i$, the expected count is $N_i=R_i{T}_i$ and its standard deviation is $\sqrt{N_i}$. The fractional rate uncertainty is consequentlyA target of $0.15$ requires $N_i\simeq (0.15{)}^{-2}=44.44$ counts at each proper-time setting. The exposure time for a Poisson rate at fixed fractional precision is therefore $T_i=1/(0.15^2{R}_i)$, givingSince the settings are acquired independently rather than simultaneously,Nearly three quarters of the exposure is spent at the latest, lowest-rate setting. Proper time in picoseconds labels the decay distribution; it is not the laboratory acquisition time in days. This estimate assumes counting statistics dominate, with no additional background-subtraction uncertainty or acquisition overhead.

The first-order grating equation is $d\sin \theta =\lambda$, so $d=2.879\mu \mathrm{m}$, about $347$ lines per millimetre. For $N$ equally illuminated lines,Principal maxima stay at the same angles, their peaks scale as $N^2$ for fixed illumination per line, and their angular widths shrink as $1/N$. The resolving power of a diffraction grating is $\lambda /\Delta \lambda =mN$. Thus first order needs $N=1000$ and illuminated width $Nd\simeq 2.88\mathrm{mm}$.

The grating dispersion follows by differentiation: $d\theta /d\lambda =1/(d\cos \theta )$. With a detector centred on the chief ray, one pixel corresponds approximately toThis exceeds the nominal $0.5\mathrm{nm}$ grating width. The practical resolution scale is therefore a few nanometres; two-pixel sampling gives about $4.5\mathrm{nm}$. Exact resolvability also depends on slit width and image sampling. Taking the detector coordinate as $f\tan \theta$ about the original axis instead changes the single-pixel estimate only to $2.20\mathrm{nm}$.

For ideal photon counting, Poisson distribution statistics give SNR $=\sqrt{N_{\gamma }}$. Hence $N_{\gamma }=10^4$ and $t_{\min }=10^4/(2\times 10^5)=0.050\mathrm{s}$.

Use a lock-in amplifier for phase-sensitive detection: chop or otherwise modulate the source before the entrance slit, retaining a synchronous reference. At each grating setting amplify the photodiode output, multiply it by a reference and low-pass filter. For a signal $s\cos ({\omega }_{r}t+\varphi )$ and reference $2\cos {\omega }_{r}t$, the low-pass output is $s\cos \varphi$; quadrature demodulation also recovers $s\sin \varphi$. Choose the modulation frequency above slow drift while within the source/detector bandwidth and away from strong electrical interference. Average for many cycles before changing the grating setting.

This rejects steady backgrounds and reduces low-frequency noise through narrow-band detection. It does not remove random dark-current/stray-light shot noise within that bandwidth, nor a background modulated coherently with the source. The chopping geometry must therefore distinguish wanted light from stray paths.^[ref]

Assume independent arrivals at constant rate $r=2000{\mathrm{s}}^{-1}$, so the count follows a Poisson distribution. In $1\mathrm{ms}$ its mean is $\mu =2$. ThereforeOver observation time $T$, the rate estimate is $\hat{r}=N/T$ and its standard deviation is $\sqrt{r/T}$. Thus the fractional one-standard-error uncertainty is $1/\sqrt{rT}$. To make this strictly less than ten per cent requiresThe mean current magnitude is $er\simeq 3.20\times 10^{-16}\mathrm{A}$. The accuracy statement needs a confidence convention: the boxed time uses one standard error, not a guaranteed bound on every realization. A Gaussian 95-percent relative interval would instead require approximately $(1.96/0.1{)}^2$ events, or $T\gtrsim 0.192\mathrm{s}$. Dead time and correlated arrivals would need another count model; the mean rate alone does not establish Poisson statistics.

Using the probability mass function of the Poisson distribution and the power series of the exponential function,This moment-generating function of a Poisson distribution is finite for every real $t$. It also covers $\mu =0$, when $X=0$ with probability one and the moment-generating function is identically one.