MyIQtested research data: IQ test scores by device (device_os)
Data set version 2026-09-29.1. Tests finished 2026-09-17 to 2026-09-29 (UTC). Generated 2026-09-30.
Page: https://www.myiqtested.com/research/iphone-vs-android-iq
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Credit MyIQtested and link to the page above.

WHAT THIS IS
Aggregates of 185649 first attempts at the MyIQtested IQ test, grouped by the device the test was taken on. Nobody was asked: the device group comes from the operating system the browser reports in its user-agent header, read by our analytics endpoint. No individual responses, visitor ids, user-agent strings or timestamps are included.
The sample is self-selected online test takers. It describes patterns in this sample, not why they occur.

DEVICE GROUPS
  iphone      iOS (iPhone and iPod)
  android     Android (phones and tablets)
  windows     Windows
  mac_ipad    macOS, plus iPadOS where the browser says so. iPadOS Safari sends a Mac user agent by default, so iPads and Macs cannot be told apart and are one group.
  chromebook  ChromeOS
Left out: Linux (1768 first attempts; most carry the country mix of Android traffic and are likely phones in desktop mode), other and unknown systems.
In-app browsers (Facebook, Instagram, TikTok and others) are kept in the group of the device they run on.

WHICH TESTS COUNT
    324088  test completions (events, one per event id) finished 2026-09-17 to 2026-09-29
    205656  ... and first completion of the test by that visitor
    205656  ... and not a preview
    205656  ... and all 33 puzzles delivered and answered
    205656  ... and untimed mode
    205656  ... and a score on the scale
    205604  ... and a country
    187421  ... and 3 to 60 minutes from start to finish
    185649  ... and one of the five device groups

FILES
  device_os-aggregates.csv    every device x age band x country cell (one header row)
  device_os-summary.csv       adults 18+ and all ages, adjusted: reproduces every number quoted on the page
  device_os-datapackage.json  Frictionless data package describing both CSVs

COLUMNS OF device_os-aggregates.csv
  site                   string   myiq (MyIQtested test takers)
  answer_id              string   machine id of the device group: iphone, android, windows, mac_ipad, chromebook
  answer_label           string   the device group as shown on the page
  age_band               string   under_13, 13_15, 16_17, 18_24, 25_34, 35_44, 45_54, 55_64, 65_plus, or unknown (prefer not to say, skipped, missing)
  country_code           string   ISO 3166 alpha-2 code, or OTHER (countries with fewer than 750 first attempts in the sample)
  country_name           string   English name of the country
  n                      integer  people in the cell
  suppressed             integer  1 when n is under 5: the statistics are left blank
  iq_gap_vs_age_country  number   mean test score of the cell minus the mean of everyone in the sample in the same age band and country, in points (score steps are 3 points = 1 question). Relative only: the scoring is a fixed conversion, not an empirical norm, so no absolute level is published
  iq_sd                  number   sample standard deviation of the scores in the cell

Cells with fewer than 5 people are suppressed (statistics left blank, suppressed = 1; n is still shown). Countries with fewer than 750 first attempts are pooled into OTHER.
The score scale is a fixed mapping (score = 51 + 3 x correct answers, range 55-155), not an empirical norm, and online takers score high on it. Differences between groups are meaningful; absolute levels are not.

COLUMNS OF device_os-summary.csv
  dataset_version        string   version of the data set these rows come from
  question_id            string   id of the data set (device_os)
  view                   string   adults_adjusted = adults 18+, all countries, adjusted for age band and country; all_ages_adjusted = everyone, adjusted
  answer_id              string   machine id of the device group
  answer_label           string   the device group as shown on the page
  n                      integer  people in this device group in the view
  share_of_view          number   n divided by everyone in the view (0 to 1)
  gap_points             number   difference in test-score points between this device group and everyone in the view (0 = same as everyone). Blank when n_too_small = 1
  ci95_low               number   lower end of the 95% confidence interval of gap_points
  ci95_high              number   upper end of the 95% confidence interval of gap_points
  mean_percentile        number   average percentile (0-100) of the group, measured against everyone of the same age band and country
  median_percentile      number   median percentile, same measure
  n_too_small            integer  1 when the group has fewer than 30 people in the view: no gap or percentile is given

HOW THE GAPS ARE COMPUTED
Adjusted views use direct standardisation: each device group is compared with everyone, within each age band x country stratum, and the stratum differences are averaged with weights proportional to the stratum sizes of everyone in the view (strata where the group has nobody are dropped and the weights renormalised). The 95% interval comes from the stratified standard error. Percentiles are measured within each person's own age band and country.
Every range is a 95% confidence interval, and every difference we point out holds at 95% confidence after correcting for the number of comparisons (Holm method).