• @uis
    link
    English
    211 months ago

    How can standard deviation be negative?

    • @[email protected]
      link
      fedilink
      English
      711 months ago

      It’s deviations from the mean, so if the deviation were “3” for example, values of 6, 3, 0, -3 and -6 would be 2, 1, 0, -1, and -2 deviations away from the central line, respectively.

      • @uis
        link
        English
        0
        edit-2
        11 months ago

        Deviation != standard deviation

        Standard deviation is square root of sums of squared deviations divided by number of samples. Only complex numbers can result in negative values when squared. Negative amout of samples makes even less sense.

        Deviation from mean is x - μ, standard deviation is this abomination:

        • @jj4211
          link
          English
          911 months ago

          The standard deviation is not negative, that data was just that many standard deviations below the mean. Think “this data point is below the mean by 0.5 standard deviations” not “the standard deviation is 0.5”. They are using standard deviation as a unit rather than, say, degrees Celsius.

          • @uis
            link
            English
            111 months ago

            Then why yellow line doesn’t touch time axis? Function cannot always be bigger than its own mean. If there is point above mean, than there should be at least one point below mean. I’m assuming here mean is of temerature in that year.

            • @jj4211
              link
              English
              511 months ago

              The chart could stand for some clarification, but it looks like the mean and standard deviation refer to statistics covering all the years from 1982 through 2011. However, it does not explicitly state the dataset over which the standard deviation is calculated, but it seems reasonable to assume that the same aggregate cited for the mean is also the same aggregate used for the standard deviation.

              Each line in the graph represents a single year of data. It’s kind of messy and only two of the years are actually labeled, 2023 and the partial data for 2024. So that bottom-most line represents some unspecified year that was consistently 1.5 to 2 standard deviations below the mean for the 30 year analysis.

              The data is at https://climatereanalyzer.org/clim/sst_daily/json/oisst2.1_world2_sst_day.json, but alas, I’m too lazy to try to reproduce this sort of analysis to verify my guesses.

              I will say it’s a peculiar approach and visualization. Including a subset of the data in the mean/standard deviation and then plotting the entire data. Also impossibly jumbled line graph visualizations of most years instead of something easier. I’d imagine you could convey the point with each year consolidated to a single data point and have a much easier to follow graph.

              • Gorgritch_Umie_Killa
                link
                fedilink
                English
                211 months ago

                So, a little while ago climate change deniers used the fact of fluctuations in temperature throughout the year as a basis for a false claim that climate scientists were hiding the ‘real’ data in the less jumbly plots you suggest the use of. (And any sensible person would see the benefits of).

                Whoever produced this is likely aware of those cynical and false claims, and decided they don’t want any risk the point they are making, being similarly undermined.

              • @uis
                link
                English
                111 months ago

                Thanks

    • @jj4211
      link
      English
      711 months ago

      It isn’t, but some data are a negative multiple of standard deviation away from the mean.

      • @uis
        link
        English
        211 months ago

        Same question as in another subthread