Terms
0-9
95% interval #
The range that shows the statistical uncertainty in a share among the analysed posts. A wide interval means less certainty, usually because there are few posts. It does not cover every source of error.
In Public Mood analyses: every weekly, platform and period share is shown with one, and a difference between two periods is shown in percentage points with its own 95% interval. In the sample analysis, the critical share from the week starting 21 Sep 2026 was 69% (95% interval 64-73%). Every analysis has sources of error, and we always do our best to state them. The 95% interval shows the statistical uncertainty in a share, but not every source of error, so read weekly movements of a few points as noise, not news.
A
Adjusted for comparisons #
Adjusted for comparisons means the bar for calling a difference real is raised when many things are compared at once, because some differences look real by chance. Holm's step-down method is one standard way to do it.
In Public Mood analyses: when several platforms, weeks, themes or periods are compared at once, we use Holm's method, which keeps the chance of at least one false alarm at 5%. We do not highlight the single smallest p-value among many. See also "Multiple comparisons".
Analysed posts #
Analysed posts are the individual posts, comments or replies that were actually counted in a result, not every post that exists about a topic.
In Public Mood analyses: every analysis states how many posts were analysed per platform, and the scope per platform. Numbers are written as "69% of stance-taking posts were critical", never "69% of people think". The sample analysis rests on 2,060 analysed posts. More in how Public Mood analyses online sentiment.
Analysis period #
The analysis period is the span of dates that a set of results covers.
In Public Mood analyses: the default is the last 30 days or from the event onwards, up to 12 months back on request, and the period is always stated. The latest week can be partial and is then flagged.
Anonymisation #
Anonymisation means removing or hiding the details that identify a person, so that a quote or a count cannot be traced back to them.
In Public Mood analyses: quotes are shown with platform and date only, never with usernames of private individuals, and ordinary users appear only as counts per platform and week. See also "Data retention".
B
Baseline (usual tone) #
A baseline is the usual level a result is read against, for example a topic's own earlier period. A figure such as 70% critical means little until it is compared with the baseline.
In Public Mood analyses: sentiment is never read as an absolute score. We read it against how the topic is normally discussed online: the change over time, the difference between groups and platforms, and, where possible, the same topic earlier or similar topics. Each platform is read against its own usual tone. A baseline across topics does not exist yet; today the baseline is the topic's own earlier period. In the sample analysis, the baseline was 31% critical (95% interval 25-38%) across the nine weeks starting 13 Jul to 7 Sep 2026.
Before-and-after comparison #
A before-and-after comparison measures a result in a period before a date or event and in a period after it, and looks at the difference. When the comparison is built around a specific event, it is often called an event study.
In Public Mood analyses: we compare the latest period with the topic's own earlier period and give the difference in percentage points with a 95% interval. The date of a shift is an estimate, with an uncertainty of about a week, and events near it are context, not proof of cause.
Bias #
Bias is a systematic error that pushes results in one direction. Unlike random error, it does not shrink when more data is added.
In Public Mood analyses: the main bias is self-selection (see "Selection bias"). We describe it and do not claim to correct for it.
Bots and coordinated activity #
Bots are automated accounts, and coordinated activity is many accounts posting in step to push a message. Both can make a view look more common online than it is.
In Public Mood analyses: some accounts may be bots, and we do not claim that bots, spam or coordinated accounts are fully excluded. We state this as a limit.
C
Clear change #
A clear change is a difference between two periods that is large enough to matter, rests on enough posts and is unlikely to be chance.
In Public Mood analyses: we call a change clear only when it is at least 10 percentage points, its 95% interval does not include 0, and each period has at least 100 stance-taking posts. Otherwise we write "no clear change" and show the numbers. We never call a change on one week alone, and a clear change shows that the mood moved, not why. In the sample analysis, the rise from 31% to 69% critical (+37 percentage points, 95% interval 29 to 45, on 201 and 338 stance-taking posts) is a clear change; the latest week against the week before (67% against 73%) is not.
Confidence interval #
A confidence interval is a range around an estimate that shows its statistical uncertainty. It is built by a method that would cover the true value in a stated share of repeated studies, usually 95%, so it does not mean there is a 95% chance that one particular range is right. It does not cover every source of error.
In Public Mood analyses: we call ours a "95% interval" and show it around shares among analysed posts. Every analysis has sources of error, and we always do our best to state them.
Correlation and causation #
Correlation means two things move together. Causation means one makes the other happen, and correlation alone does not show it.
In Public Mood analyses: a clear change shows that the mood moved, not why. Matches, speeches or headlines listed near a shift are context, not proof of cause.
Coverage #
Coverage has two meanings in statistics: how well a list includes everything it should, and how often an interval method contains the true value.
In Public Mood analyses: "news coverage" means written online news articles on the topic. When your topic is in the news, we show the coverage week by week next to the online mood, so you can see how the two moved together. The outlet count is based on the sources we can find, so it is a minimum, not a complete list. TV, radio, podcasts and print-only editions are not included.
Critical post #
A critical post is a post that takes a side against the topic, for example by criticising a person, decision or product.
In Public Mood analyses: critical is one of our three labels, with supportive and neutral. In the sample analysis, 328 of 2,060 analysed posts were critical.
Critical share #
The critical share is the share of stance-taking posts that are critical: critical posts divided by critical plus supportive posts.
In Public Mood analyses: it is the headline measure. Neutral posts are left out of both numerator and denominator, so it is never the share of all analysed posts. In the sample analysis, 328 of 607 stance-taking posts were critical (54%), while critical posts were only 16% of all 2,060 analysed posts.
D
Data retention #
Data retention is how long data is kept, and what happens to it afterwards.
In Public Mood analyses: raw posts are deleted 90 days after delivery. Only the anonymised results are kept.
Digital fingerprint #
A short code calculated from a file. Change a single character in the file and the code is completely different, so the same code means the same file.
In Public Mood analyses: every analysis we deliver gets a SHA-256 fingerprint of its PDF, its data file and a locked copy of its report page, listed in our public register.
E
Echo chamber #
An echo chamber is a community or feed where similar views are repeated and opposing views are rare, so the conversation there leans one way.
In Public Mood analyses: the communities and channels in scope shape the result, so the scope per platform is always stated. In the sample analysis, Reddit covered one community, r/reddevils.
Engagement #
Engagement is how much interaction a post gets, such as likes, shares, replies or views. A post with high engagement is seen by many, but engagement does not show how common a view is.
In Public Mood analyses: every percentage we report counts each analysed post once, whatever its likes or reach, and quotes are chosen by a fixed rule, not by likes or reach.
I
Illustrative quote #
An illustrative quote is a short excerpt shown to give a sense of what a group of posts says. It illustrates a result and does not prove it.
In Public Mood analyses: a quote is shown only when the analysis measures the mood it expresses, and next to it we state how large that group is. We show at most three short, unchanged, anonymous quotes, one supportive, one neutral and one critical, each chosen by a fixed rule, never by likes or reach, and read by hand. Quotes are shown with platform and date only, with no names and no links.
Independent observations #
Observations are independent when knowing one tells you nothing about another. Posts in the same thread, video or by the same account usually are not, so the effective number of observations is smaller than the count.
In Public Mood analyses: our intervals and tests treat posts as independent and are therefore likely too narrow. We do not claim to correct for this; reports show how a main comparison holds up if posts were ten times less independent.
M
Margin of error #
A margin of error is the plus-or-minus range reported for a result when the people behind it are a cross-section of everyone, chosen so that each had a known chance of being included. It shows the statistical uncertainty that comes from observing some members of a population instead of all of them. It does not cover every source of error, such as question wording or who declined to take part.
In Public Mood analyses: we give none. Online posts come from people who chose to post, so a margin of error would be misleading. Instead every analysis states the posts per platform, the period and the limits, and shows 95% intervals for shares among the analysed posts. The two statements do not contradict each other. Every analysis has sources of error, and we always do our best to state them.
The media pickup check re-checks, for a limited time, whether other outlets have cited or picked up a published article after publication.
In Public Mood analyses: customers can add their own published articles and we re-check for up to 31 days, reported as "cited by at least X outlets". It does not update the analysis and is not live monitoring.
Mood (online mood) #
Online mood is the balance of critical, supportive and neutral posts about a topic at a given time, and how it shifts.
In Public Mood analyses: "online mood" is the plain-English word for online sentiment. Both describe analysed posts from accounts that chose to post publicly.
Multiple comparisons #
The multiple comparisons problem is that the more differences you test at once, the more likely it is that some look real by chance alone. Methods that adjust for it raise the bar for each single comparison.
In Public Mood analyses: we adjust with Holm's method (see "Adjusted for comparisons"). In the sample analysis, X was clearly less critical than Reddit and YouTube after adjustment (43% against 61% on YouTube), while Reddit and YouTube did not differ clearly.
N
Neutral post #
A neutral post is a post with no clear stance towards the topic. It is not counted in the shares.
In Public Mood analyses: the share of neutral posts differs a lot between platforms (in the sample analysis 47% on X, 78% on Reddit, 77% on YouTube). A difference in neutral share is not a difference in sentiment.
Non-response bias #
Non-response bias arises when those who do not take part differ from those who do, so the results lean towards those who did.
O
Online mood reading #
An online mood reading is a description of how a topic is discussed in public online posts over a stated period: which stances are taken, how strongly and how that changes. It describes posts, not what a population thinks.
In Public Mood analyses: this is what every analysis is, and how we ask results to be described when they are published, together with the period, platforms and number of posts analysed.
Outlier #
An outlier is a value far from the rest of the data, such as a day with an unusually high number of posts or one account that posts constantly. Outliers can distort averages and shares.
P
p-value #
A p-value is the probability of seeing a difference at least as large as the one observed if there were really no difference. It is not the probability that a result is true, and it says nothing about how big or important a difference is.
In Public Mood analyses: we report the size of a difference in percentage points first, with its interval. The p-value is secondary, and we do not show a weekly "significant" flag.
Percentage point #
A percentage point is the unit for the difference between two percentages. Going from 30% to 50% is 20 percentage points, not 67%.
In Public Mood analyses: all changes are given in percentage points, for example "+37 percentage points (95% interval +29 to +45)" in the sample analysis.
The platform mix is the share of analysed posts that comes from each platform.
In Public Mood analyses: a platform with many posts weighs more in a pooled share, so the platform split is always shown and the mix must be read together with the share. Truth Social is shown separately and never pooled with other platforms.
Pooling #
Pooling means adding the counts of several weeks, groups or platforms together to get one figure that rests on more observations.
In Public Mood analyses: thin weeks are pooled with neighbouring weeks, and a pooled figure is always stated as pooled. In the sample analysis, single weeks in July to early September rested on 17 to 32 stance-taking posts each and ranged from 14% to 47% critical, so the nine weeks were pooled into one baseline of 31% (63 of 201).
Population #
The population is the whole group a study wants to say something about, for example all adults in a country.
In Public Mood analyses: the numbers describe analysed posts, not a population.
Practical significance #
Practical significance is whether a difference is large enough to matter, regardless of whether it is statistically significant. With enough data, even a tiny difference can be statistically significant without being important.
In Public Mood analyses: a clear change needs at least 10 percentage points, so that small differences in large topics do not become headlines.
Probability sample #
A probability sample is a group chosen so that every member of the population has a known chance of being picked. It is what makes a margin of error and statements about the whole population possible.
In Public Mood analyses: our analysed posts are not a cross-section of anyone. They come from accounts that chose to post publicly, which is why we do not report a margin of error and why a Public Mood analysis is an online mood reading, not representative research.
Public figures are people who hold a public role, such as politicians, officials, celebrities or club managers. Private individuals are everyone else who posts online.
In Public Mood analyses: private individuals are never named or linked. We only link to news outlets, organisations, public figures and public communities and channels, and ordinary users appear only as counts per platform and week.
R
Reporting week #
A reporting week is the fixed seven-day window that counts are grouped into, so that weeks can be compared on the same terms.
In Public Mood analyses: a week runs Monday to Sunday and is labelled by its Monday. The latest week can be partial and is then flagged; in the sample analysis, the final week covers only 12-13 Oct 2026.
Representative sample #
A representative sample is a sample whose make-up mirrors the population on the traits that matter, such as age, region or gender. Random selection and weighting are the usual ways to get there.
In Public Mood analyses: the analysed posts are not representative of a population, voters, customers or fans, and we do not present them as such.
Response rate #
The response rate is the share of those asked to take part who actually do. A low rate raises the risk that those who answered differ from those who did not.
In Public Mood analyses: there is no response rate, because nobody is asked. Posts are written on the poster's own initiative.
S
Sample #
A sample is the part of a population that is actually observed, used to draw conclusions about the whole.
In Public Mood analyses: we avoid calling our posts "a sample", because they are not drawn from a population; we say "analysed posts". "Sample analysis" on our site means an example report, not a statistical sample.
Sample size #
Sample size, written n, is the number of observations a result rests on. More observations give a narrower range of uncertainty, but do not remove bias.
In Public Mood analyses: the count that matters is the number of stance-taking posts, because neutral posts do not count towards a share. Weeks with fewer than 30 stance-taking posts get an interval and no verdict, and a clear change needs at least 100 in each period. An analysis covers up to 10,000 posts.
Sampling error #
Sampling error is the difference between a result from a sample and the true value for the whole population that arises because only part of the population was observed.
In Public Mood analyses: we do not report sampling error, because our posts come from people who chose to post, not a cross-section of everyone. The 95% interval shows the statistical uncertainty in a share among the analysed posts only.
Search interest (Google Trends index) #
Search interest is how often a term is searched on Google, shown by Google Trends as a relative index where 100 is the peak for that term and region.
In Public Mood analyses: it shows attention, not sentiment and not search counts. It is shown beside the sentiment and never mixed into the shares.
Seasonality #
Seasonality is a pattern that repeats at regular times, such as weekends, holidays or the start and end of a sports season. It can look like a change when it is only the usual rhythm.
Selection bias (self-selection) #
Selection bias arises when what is observed is not a fair picture of the whole, for example because it chose itself. Self-selection is selection bias in which participants decide to take part. Sampling bias is the related term for a selection method that favours some members of a population over others.
In Public Mood analyses: those who post differ from those who do not, some groups post far more than others, and a few very active accounts can shape the conversation. Platform audiences and the communities or videos in scope also shape the result.
Sentiment (online sentiment) #
Sentiment is the stance or feeling a post expresses towards a topic. Online sentiment is the balance of those stances across many public posts.
In Public Mood analyses: each analysed post is labelled critical, supportive or neutral towards the topic, and "online sentiment" is the balance of those labels over time. Every analysis has sources of error, and we always do our best to state them, so read small differences with care.
Sentiment analysis #
Sentiment analysis is the measurement of the stance or feeling expressed in text, usually across many posts or articles, and of how it changes. Online sentiment analysis applies it to public online posts, and is also called social media sentiment analysis or sentiment tracking.
In Public Mood analyses: results are read against the topic's own baseline, never as an absolute score, and each analysis states its platforms, period, counts and caveats.
Share #
A share is a part expressed as a percentage of a stated whole. The whole must always be named.
In Public Mood analyses: the whole is stance-taking posts (critical plus supportive), for one platform or one period, with the number of posts behind it.
Share of voice #
Share of voice is the share of all online mentions in a market or topic that refer to one brand, person or side, compared with the others. It measures visibility, not stance.
Social listening #
Social listening is the practice of following public online conversation about a topic, brand or person, and analysing what is said and how it changes.
In Public Mood analyses: we deliver finished analyses with a verdict, written for editors, rather than software to operate yourself.
Sources tab #
A sources list shows where the material behind a result comes from, so readers can see what it rests on.
In Public Mood analyses: every analysis has a Sources tab in three tiers: news and fact-check sources, public channels with the number of posts analysed, and private users shown only as counts per platform and week. Private individuals are never named or linked.
Spike #
A spike is a short, sharp rise in the number of posts about a topic, often around an event. It shows a burst of attention, not a change of mood.
In Public Mood analyses: volume charts show how much was posted, not the posts we analysed, and a busy day is not a more critical day.
Stance #
Stance is the position a post takes towards a topic: for it, against it or neither.
In Public Mood analyses: stance is recorded with three labels: supportive (for), critical (against) and neutral (no clear side).
Stance-taking post #
A stance-taking post is a post that clearly takes a side, for example for or against. Neutral and unclear posts are not counted in shares.
In Public Mood analyses: in the sample analysis, 607 of 2,060 analysed posts took a side.
Standard deviation #
Standard deviation is a measure of how spread out values are around their average. About two thirds of values in a bell-shaped distribution lie within one standard deviation of the average.
Statistical significance #
A result is statistically significant when it would be unlikely to occur by chance alone if there were no real effect, judged against a chosen threshold such as 5%. It does not mean the result is large, important or caused by anything in particular (see "Practical significance").
In Public Mood analyses: we do not call results "significant" or "statistically proven" without the size and interval. We use the clear-change rule, which combines a minimum size, an interval and a minimum number of posts, and is a decision rule, not a significance claim.
Supportive post #
A supportive post is a post that takes a side for the topic, for example by backing a person, decision or product.
In Public Mood analyses: supportive is one of our three labels, with critical and neutral. In the sample analysis, 279 of 2,060 analysed posts were supportive.
T
Thin week #
A thin week is a week with so few relevant observations that its figure mostly shows chance variation.
In Public Mood analyses: weeks with fewer than 30 stance-taking posts get an interval and no verdict, and may be pooled with neighbouring weeks. In the sample analysis, the week of 28 Sep 2026 had only 19 stance-taking posts (95% interval 41-81%), so it is shown without a verdict.
Timestamp #
A signed receipt from an independent service showing that a digital fingerprint existed at a given time.
In Public Mood analyses: each analysis is timestamped at delivery by two independent services, normally DigiCert and Sectigo, under the open RFC 3161 standard. It proves the analysis is unchanged since delivery, not that it is right.
Transition week #
A transition week is a week whose figure lies between the level before a shift and the level after it.
In Public Mood analyses: the week just before a dated shift is shown separately when its share lies between the two levels. In the sample analysis, the week starting 14 Sep 2026 was a transition week (49% critical, 68 stance-taking posts).
Trend #
A trend is a lasting direction of change over time, as opposed to short-term ups and downs.
In Public Mood analyses: we avoid trend inference from short weekly series. We compare two defined periods, date the change as an estimate with an uncertainty of about a week, and do not forecast.
V
Volume #
Volume is the number of posts about a topic in a period.
In Public Mood analyses: volume shows how much is being said, not whether the mood changed. Volume charts are not the posts we analysed, and a busy day is not a more critical day.
W
Weighting #
Weighting means giving some observations more influence than others so that the observed group matches the population on known traits, such as age or region.
In Public Mood analyses: results are not weighted to any population. Shares are plain counts of analysed posts.
Wilson interval #
The Wilson interval is a 95% interval for a single share. It stays between 0% and 100% and works for small counts, where simpler formulas fail.
In Public Mood analyses: it is the interval we use around every single share (weekly, platform or period).
