autocorrelation
C2Technical/Academic/Formal
Definition
Meaning
A statistical measure of how a signal or time series is correlated with a delayed copy of itself; self-similarity across different points in time.
In various fields, the degree to which a process or set of data points resembles itself over successive intervals, often used to detect patterns, periodicities, or randomness. In finance, it can indicate market inefficiency; in signal processing, it helps analyze frequency components.
Linguistics
Semantic Notes
The term is strictly technical and carries no figurative or informal meaning. It implies a mathematical or computational operation. Often associated with concepts like lags, time series, and stochastic processes.
Dialectal Variation
British vs American Usage
Differences
No significant lexical or spelling differences; the concept and term are identical in both variants. Minor pronunciation differences as captured in IPA.
Connotations
Identical technical connotations in both dialects.
Frequency
Equally low frequency in general discourse but standard in technical contexts (e.g., statistics, econometrics, signal processing) in both regions.
Vocabulary
Collocations
Grammar
Valency Patterns
The autocorrelation of [NOUN PHRASE] was [ADJECTIVE]To compute/calculate/find the autocorrelationAutocorrelation at lag [NUMBER]Vocabulary
Synonyms
Strong
Neutral
Weak
Vocabulary
Antonyms
Usage
Context Usage
Business
Used in econometrics and financial analysis to detect patterns in stock returns or economic indicators over time.
Academic
A core concept in statistics, signal processing, time series analysis, and spatial statistics.
Everyday
Virtually never used in everyday conversation.
Technical
Precise mathematical term for analysing signals, data sequences, or spatial patterns for self-similarity.
Examples
By Part of Speech
verb
British English
- The data was autocorrelated to identify periodic trends.
- We need to autocorrelate the signal before filtering.
American English
- The time series was autocorrelated to check for seasonality.
- The algorithm autocorrelates the input to detect repeating patterns.
Examples
By CEFR Level
- Scientists use autocorrelation to find repeating patterns in climate data.
- If the autocorrelation is high, yesterday's temperature is a good predictor of today's.
- A significant positive autocorrelation at lag one in the financial returns indicates a trend-following market behaviour.
- The geostatistical model accounted for spatial autocorrelation to avoid biased estimates.
Learning
Memory Aids
Mnemonic
Think of a hallway of mirrors (AUTO) where each reflection is a slightly delayed version of yourself. The strength of the resemblance between you and your delayed reflection is the CORRELATION. Hence, AUTO-CORRELATION.
Conceptual Metaphor
A signal's MEMORY of itself; the ECHO of a process; a data set looking in a (temporal) MIRROR.
Watch out
Common Pitfalls
Translation Traps (for Russian speakers)
- Do not translate as 'автокорреляция' in non-technical contexts, as it is a direct loanword and will be misunderstood.
- Avoid confusing with 'автокоррекция' (auto-correction).
- The prefix 'auto-' here means 'self', not 'automatic'.
Common Mistakes
- Misspelling as 'auto-correlation' (hyphen is generally omitted in modern technical writing).
- Confusing it with cross-correlation (which involves two different signals).
- Using it as a verb (e.g., 'The data autocorrelates' is non-standard; prefer 'exhibits autocorrelation' or 'is autocorrelated').
Practice
Quiz
In which field is autocorrelation LEAST likely to be a standard analytical tool?
FAQ
Frequently Asked Questions
It is neither inherently good nor bad; it is a property of data. In statistical modelling (like regression), unexpected autocorrelation in residuals is problematic as it violates the independence assumption. In signal analysis, it is a useful tool for finding hidden periodicities.
Autocovariance is the raw covariance of a signal with its lagged self. Autocorrelation is the normalised version (typically divided by the variance), yielding a value between -1 and 1, making it scale-independent.
A zero autocorrelation at a specific lag means there is no linear relationship between the signal's current value and its value at that lag. If autocorrelation is zero at all lags (except lag 0), the process is called white noise.
Yes. Negative autocorrelation means that a high value is likely to be followed by a low value, and vice versa (e.g., mean-reverting processes).