Research talk:Reading time/Work log/2018-09-27
Latest comment: 6 years ago by Tbayer (WMF) in topic Question: General patterns of time spent
Thursday, September 27, 2018
editQuestion: General patterns of time spent
editWhat are the general patterns of time spent? Calculate averages and various percentiles, characterize the form of the distribution / histogram (cf. https://www.nngroup.com/articles/how-long-do-users-stay-on-web-pages/ )
Summary stats
editVisible Length
edit| wiki | max | percentile_95 | percentile_75 | median | mean | percentile_25 | percentile_5 | min |
|---|---|---|---|---|---|---|---|---|
| arwiki | 1.537986e+12 | 373673.4 | 71098.0 | 25296.0 | 7.035699e+08 | 8979.25 | 2402.0 | -1.724141e+06 |
| dewiki | 1.537995e+12 | 409897.0 | 74609.5 | 23406.5 | 1.448944e+08 | 7397.00 | 1861.0 | -6.982139e+06 |
| enwiki | 1.537952e+12 | 392193.7 | 61535.0 | 19111.0 | 1.184951e+09 | 6282.50 | 1521.0 | -1.594799e+09 |
Total Length
edit| wiki | max | percentile_95 | percentile_75 | median | mean | percentile_25 | percentile_5 | min |
|---|---|---|---|---|---|---|---|---|
| arwiki | 1.537986e+12 | 794130.30 | 91225.5 | 28504.0 | 7.039803e+08 | 9555.00 | 2490.0 | -1.724141e+06 |
| dewiki | 1.537995e+12 | 1199123.75 | 99507.0 | 26876.5 | 1.453318e+08 | 7828.75 | 1912.0 | -6.982139e+06 |
| enwiki | 1.537952e+12 | 1105116.50 | 86771.0 | 22400.0 | 1.185385e+09 | 6711.00 | 1585.0 | -1.594799e+09 |
Dom Interactive Time
edit| wiki | max | percentile_95 | percentile_75 | median | mean | percentile_25 | percentile_5 | min |
|---|---|---|---|---|---|---|---|---|
| arwiki | 293013 | 8248.25 | 2380 | 1261 | -785805047 | 719 | 254 | -1537985844507 |
| dewiki | 7808564 | 2686.25 | 934 | 516 | -216542229 | 285 | 115 | -1537995044942 |
| enwiki | 676221 | 4646.90 | 1496 | 765 | -1239425778 | 386 | 128 | -1538004439131 |
There are negative values so lets see what it looks like when we forcibly remove them
editVisible Length
edit| wiki | max | percentile_95 | percentile_75 | median | mean | percentile_25 | percentile_5 | min |
|---|---|---|---|---|---|---|---|---|
| arwiki | 1.537986e+12 | 373753.50 | 71102.50 | 25300.5 | 7.036840e+08 | 8984.25 | 2403.85 | 17.0 |
| dewiki | 1.537995e+12 | 410429.00 | 74619.00 | 23417.0 | 1.449385e+08 | 7405.00 | 1866.80 | 10.0 |
| enwiki | 1.537952e+12 | 392454.35 | 61576.25 | 19126.0 | 1.185528e+09 | 6293.00 | 1532.00 | 2.0 |
Total Length
edit| wiki | max | percentile_95 | percentile_75 | median | mean | percentile_25 | percentile_5 | min |
|---|---|---|---|---|---|---|---|---|
| arwiki | 1.537986e+12 | 794295.0 | 91230.0 | 28508.0 | 7.040374e+08 | 9556.0 | 2491.0 | 17.0 |
| dewiki | 1.537995e+12 | 1200417.1 | 99588.5 | 26883.5 | 1.453691e+08 | 7837.0 | 1914.0 | 10.0 |
| enwiki | 1.537952e+12 | 1106109.5 | 86814.5 | 22412.5 | 1.185687e+09 | 6715.0 | 1589.0 | 2.0 |
Dom Interactive Time
edit| wiki | max | percentile_95 | percentile_75 | median | mean | percentile_25 | percentile_5 | min |
|---|---|---|---|---|---|---|---|---|
| arwiki | 293013 | 8251.55 | 2381 | 1262 | 2531.874689 | 720 | 256 | 13 |
| dewiki | 7808564 | 2687.20 | 934 | 516 | 1128.534421 | 285 | 116 | 10 |
| enwiki | 676221 | 4660.00 | 1501 | 770 | 1574.989014 | 390 | 136 | 11 |
Notes
editThe units are milliseconds so the media time is 25 seconds on arwiki and a whole 19 seconds enwiki. So, as expected, most pageviews do not entail much reading. At the 75th percentile people are reading for 60 or 70 seconds.
Removing negative extreme values doesn't change the look of the distribution very much. One pretty striking thing is the 95th percentile for Arabic is much lower than the other two languages. This may be a sign of interesting variation between Wikis.
- @Groceryheist: Isn't the median visible time rather 25 seconds on arwiki and 19 seconds on enwiki? (The above table gives 25296.0 and 19111.0 milliseconds, respectively.) Regards, Tbayer (WMF) (talk) 16:46, 28 September 2018 (UTC)
- Yep.
Weibull Survival Curve
edit
This plot shows the weibull distribution fit to data on dwell times on wikipedia pages in 2018.. The Weibull distribution fit to dwell times of views lasting less than 10 minutes. Units on the x axis are miliseconds.
The model might not be a perfect fit for the data. There are fancier versions of the weibull distribution with some more degrees of freedom that might be worth a shot.
This plot shows the hazard function estimated using a weibull model on data on dwell times (total length) on wikipedia pages in 2018.. The hazard of leaving within the first few seconds is high, but it falls very fast. The model was only fit on views lasting less than 10 minutes.
This plot shows the survival curve estimated using a weibull model on data on dwell times on wikipedia pages in 2018.. Y is the proportion of readers who remain on the page after X seconds. It is obvious that most page views last less than 100 seconds, but the tail stays fat for some time.
Summary Stats By Wiki and Mobile
edit| totalLength | totalLength | visibleLength | visibleLength | weigh | weigh |
|---|---|---|---|---|---|
| median | mean | median | mean | median | mean |
| 28845.0 | 174915188.91779184 | 24635.0 | 174217568.52800092 | 3.2892 | 16.946110355694543 |
| 26818.0 | 1669471005.8635123 | 24569.0 | 1669411369.846813 | 7.1168 | 32.75366588732676 |
| 30106.0 | 1258528.345518783 | 25555.0 | 509133.7363905112 | 9.9118 | 23.65676761805455 |
| 31064.0 | 316705.2144088906 | 28627.0 | 85457.0216516574 | 1.8842 | 10.087100691852081 |
| 24924.0 | 871699107.4503895 | 21093.0 | 870993691.4453386 | 1.9374 | 72.85845601241515 |
| 29235.5 | 285826101.3141768 | 26676.5 | 285564072.7414244 | 4.902 | 60.06471769662422 |
| 37397.5 | 1077391.6539779217 | 30595.5 | 415022.3476398934 | 1.7234 | 11.103956684861044 |
| 31772.0 | 169485.2455791335 | 29706.5 | 79346.0541556145 | 4.5882 | 10.107531165980753 |
| 24508.0 | 4962884624.735496 | 19949.0 | 4962133335.288904 | 1.8062 | 232.03379988987842 |
| 23805.5 | 1029272503.6320872 | 21582.5 | 1029012633.7137628 | 12.9206 | 306.2518921945862 |
| 36873.5 | 969611.1817515872 | 28611.0 | 391199.4972326225 | 1.5786 | 156.972208235033 |
| 32178.5 | 737868769.1272662 | 29261.0 | 737723321.3466682 | 3.0652 | 126.17442748688784 |
| 34696.0 | 131896817.20872489 | 25135.0 | 130888897.00474179 | 1.8128 | 18.52878503565453 |
| 25063.5 | 350839730.9314433 | 22789.0 | 350770752.4947881 | 8.3394 | 30.900083648565893 |
| 18269.0 | 1028914.1437243772 | 15919.5 | 430494.5592872911 | 5.1216 | 14.597655231314327 |
| 27197.0 | 182449.98635280525 | 25138.0 | 73057.70582713553 | 5.7402 | 13.704078379772731 |
| 29930.0 | 2547262341.693453 | 24193.0 | 2546673837.2712193 | 1.0 | 88.85968339404555 |
| 30563.0 | 1055435486.4415013 | 28254.0 | 1055168398.4867338 | 3.4564 | 57.28288987118162 |
| 27179.0 | 437573.2503338594 | 23208.0 | 232124.67791671725 | 1.7824 | 3.9070226827241483 |
| 33413.0 | 1383063442.6839483 | 31199.0 | 1383022276.4208739 | 3.1334 | 16.950221095595637 |
| 32167.0 | 838633.2519352006 | 26587.0 | 417351.2963051632 | 3.0192 | 17.426829081511038 |
| 33911.0 | 265606748.73555964 | 31566.0 | 265483609.15417653 | 3.875 | 10.723606184090409 |
| 40412.0 | 1088856.4154136819 | 29860.0 | 513500.74116350466 | 1.2138 | 12.69478339303754 |
| 29192.5 | 2551820776.8959947 | 26793.5 | 2551722394.6258397 | 4.9252 | 26.20950873555077 |
| 26087.5 | -55590296798.80214 | 21529.0 | -55590691790.59683 | 1.099 | 61.91090603932537 |
| 24595.0 | 282794263.3259684 | 22976.0 | 282673868.4118517 | 4.2982 | 73.32214906833563 |
| 25657.0 | 1078906.9293046475 | 21712.0 | 546936.5728749209 | 3.5194 | 69.89366405109722 |
| 29184.0 | 222089215.81175953 | 27319.0 | 221970938.92457706 | 3.9904 | 102.71797857932317 |
| 14178.0 | 2260612.3961137356 | 12153.0 | 1531778.6595953293 | 1.5782 | 12.0786661764659 |
| 17027.0 | 517374010.6004746 | 16278.0 | 517322331.3574879 | 7.2122 | 17.735458991298245 |
| 25257.0 | 123572734.63591498 | 21509.0 | 122742816.4522009 | 25.0296 | 37.01092972669119 |
| 26420.0 | 326825.1428120811 | 24645.0 | 65972.88005677785 | 7.676 | 21.305935178682407 |
| 23360.0 | 865985.4052489576 | 19468.0 | 325259.689477557 | 1.1806 | 55.5853534467924 |
| 25949.0 | 79050885.803218 | 24109.0 | 78838592.91039942 | 4.3076 | 38.99953499100536 |
| 35174.0 | 365263228.88395035 | 27556.0 | 364746525.7690297 | 8.2934 | 70.0111864819028 |
| 32766.0 | 473100068.3315456 | 29928.0 | 472974081.4477851 | 2.2454 | 52.07376665702261 |
| 32835.0 | 839716.8144871682 | 27038.0 | 327017.3187352194 | 2.359 | 9.995905689297862 |
| 30140.0 | 176908.85000616751 | 27377.0 | 76993.60515603799 | 2.7172 | 11.109753345792715 |
| 33286.0 | -591539382.0003412 | 26094.0 | -592129348.2488116 | 1.4296 | 78.0922517455317 |
| 30791.0 | 479841783.5957664 | 28294.0 | 479721181.3019358 | 11.352 | 58.74345791733355 |
| 24364.0 | 1748140.32689424 | 20196.0 | 667188.4238334657 | 9.9306 | 21.768758776843907 |
| 26287.0 | 273812.2837800241 | 24358.0 | 61037.83798571826 | 4.5208 | 16.028976104766148 |
| 29498.0 | 710850.6424361493 | 22158.0 | 271006.72960397066 | 3.383 | 18.47230319340372 |
| 24450.0 | 276085265.8157966 | 22902.0 | 276011842.45716095 | 2.3744 | 15.159491895323391 |
| 39818.0 | 578659818.4633219 | 30615.5 | 578001268.7992525 | 3.5986 | 13.460458529013426 |
| 34323.0 | 644354896.7857627 | 31503.5 | 644234910.2572033 | 3.3344 | 9.49406665404602 |
| 40974.0 | 130539891.2907989 | 30675.0 | -607090.4977808126 | 3.5222 | 16.240576632530974 |
| 31655.0 | 795806196.8223436 | 28327.5 | 795698144.3101834 | 3.0226 | 8.986763498062633 |
| 28160.0 | 104098102.98057792 | 21564.0 | 103572079.54887779 | 1.4012 | 41.869781876235585 |
| 28228.0 | 809934104.2042571 | 24836.0 | 809777827.4815011 | 4.2268 | 30.325948931943454 |
Things I want to find out
edit- What is the proportion of mobile edits to each language edition and how does this compare to readership? - Does mobile editing moderate the gap between mobile and non-mobile editors?