programming - r studio ggplot download




Zeichnen Sie mit 2 y-Achsen, einer y-Achse auf der linken Seite und einer anderen y-Achse auf der rechten Seite (8)

Ich muss ein Balkendiagramm mit Zählern und ein Liniendiagramm zeigen, das alle in einem Diagramm geom_bar , ich kann beide getrennt machen, aber wenn ich sie zusammensetze, wird die erste Ebene (dh die geom_bar ) von der Skala überlagert zweite Schicht (dh die geom_line ).

Kann ich die Achse der geom_line nach rechts verschieben?


Beginnend mit ggplot2 2.2.0 können Sie eine sekundäre Achse wie folgt hinzufügen (aus der Ankündigung von ggplot2 2.2.0 ):

ggplot(mpg, aes(displ, hwy)) + 
  geom_point() + 
  scale_y_continuous(
    "mpg (US)", 
    sec.axis = sec_axis(~ . * 1.20, name = "mpg (UK)")
  )

Das ist in ggplot2 nicht möglich, weil ich glaube, dass Plots mit separaten y-Skalen (nicht y-Skalen, die sich gegenseitig transformieren) grundsätzlich fehlerhaft sind. Einige Probleme:

  • Sie sind nicht invertierbar: Wenn Sie einen Punkt auf dem Zeichenbereich angeben, können Sie ihn nicht eindeutig auf einen Punkt im Datenbereich zurückgeben.

  • Sie sind im Vergleich zu anderen Optionen relativ schwer zu lesen. Eine Studie zu Dual-Scale Data Charts von Petra Isenberg, Anastasia Bezerianos, Pierre Dragicevic und Jean-Daniel Fekete für Details.

  • Sie können leicht manipuliert werden, um in die Irre zu führen: Es gibt keine eindeutige Möglichkeit, die relativen Skalierungen der Achsen festzulegen und sie für Manipulationen offen zu lassen. Zwei Beispiele aus dem Junkcharts-Blog: one , two

  • Sie sind willkürlich: Warum haben nur 2 Skalen, nicht 3, 4 oder 10?

Vielleicht möchten Sie auch die lange Diskussion von Stephen Few zum Thema Dual-Scaled Axes in Graphen lesen. Sind sie die beste Lösung? .


Der folgende Artikel half mir, zwei von ggplot2 in einer einzelnen Zeile generierte Plots zu kombinieren:

Mehrere Grafiken auf einer Seite (ggplot2) von Cookbook for R

Und so könnte der Code in diesem Fall aussehen:

p1 <- 
  ggplot() + aes(mns)+ geom_histogram(aes(y=..density..), binwidth=0.01, colour="black", fill="white") + geom_vline(aes(xintercept=mean(mns, na.rm=T)), color="red", linetype="dashed", size=1) +  geom_density(alpha=.2)

p2 <- 
  ggplot() + aes(mns)+ geom_histogram( binwidth=0.01, colour="black", fill="white") + geom_vline(aes(xintercept=mean(mns, na.rm=T)), color="red", linetype="dashed", size=1)  

multiplot(p1,p2,cols=2)

Der schwierigste Teil war für mich, die Transformationsfunktion zwischen den beiden Achsen herauszufinden. Ich habe myCurveFit dafür verwendet.

> dput(combined_80_8192 %>% filter (time > 270, time < 280))
structure(list(run = c(268L, 268L, 268L, 268L, 268L, 268L, 268L, 
268L, 268L, 268L, 263L, 263L, 263L, 263L, 263L, 263L, 263L, 263L, 
263L, 263L, 269L, 269L, 269L, 269L, 269L, 269L, 269L, 269L, 269L, 
269L, 261L, 261L, 261L, 261L, 261L, 261L, 261L, 261L, 261L, 261L, 
267L, 267L, 267L, 267L, 267L, 267L, 267L, 267L, 267L, 267L, 265L, 
265L, 265L, 265L, 265L, 265L, 265L, 265L, 265L, 265L, 266L, 266L, 
266L, 266L, 266L, 266L, 266L, 266L, 266L, 266L, 262L, 262L, 262L, 
262L, 262L, 262L, 262L, 262L, 262L, 262L, 264L, 264L, 264L, 264L, 
264L, 264L, 264L, 264L, 264L, 264L, 260L, 260L, 260L, 260L, 260L, 
260L, 260L, 260L, 260L, 260L), repetition = c(8L, 8L, 8L, 8L, 
8L, 8L, 8L, 8L, 8L, 8L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 9L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 7L, 5L, 5L, 
5L, 5L, 5L, 5L, 5L, 5L, 5L, 5L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 6L, 
6L, 6L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 4L, 4L, 4L, 4L, 
4L, 4L, 4L, 4L, 4L, 4L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L, 0L
), module = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = "scenario.node[0].nicVLCTail.phyVLC", class = "factor"), 
    configname = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
    1L, 1L), .Label = "Road-Vlc", class = "factor"), packetByteLength = c(8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 
    8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L, 8192L
    ), numVehicles = c(2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
    2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
    2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
    2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
    2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
    2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
    2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L
    ), dDistance = c(80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L, 
    80L, 80L, 80L, 80L, 80L, 80L, 80L, 80L), time = c(270.166006903445, 
    271.173853699836, 272.175873251122, 273.177524313334, 274.182946177105, 
    275.188959464989, 276.189675339937, 277.198250244799, 278.204619457189, 
    279.212562800009, 270.164199199177, 271.168527215152, 272.173072994958, 
    273.179210429715, 274.184351047337, 275.18980754378, 276.194816792995, 
    277.198598277809, 278.202398083519, 279.210634593917, 270.210674322891, 
    271.212395107473, 272.218871923292, 273.219060500457, 274.220486359614, 
    275.22401452372, 276.229646658839, 277.231060448138, 278.240407241942, 
    279.2437126347, 270.283554249858, 271.293168593832, 272.298574288769, 
    273.304413221348, 274.306272082517, 275.309023049011, 276.317805897347, 
    277.324403550028, 278.332855848701, 279.334046374594, 270.118608539613, 
    271.127947700074, 272.133887145863, 273.135726000491, 274.135994529981, 
    275.136563912708, 276.140120735361, 277.144298344151, 278.146885137621, 
    279.147552358659, 270.206015567272, 271.214618077209, 272.216566814903, 
    273.225435592582, 274.234014573683, 275.242949179958, 276.248417809711, 
    277.248800670023, 278.249750333404, 279.252926560188, 270.217182684494, 
    271.218357511397, 272.224698488895, 273.231112784327, 274.238740508457, 
    275.242715184122, 276.249053562718, 277.250325509798, 278.258488063493, 
    279.261141590137, 270.282904173953, 271.284689544638, 272.294220723234, 
    273.299749415592, 274.30628880553, 275.312075103126, 276.31579134717, 
    277.321905523606, 278.326305136748, 279.333056502253, 270.258991527456, 
    271.260224091407, 272.270076810133, 273.27052037648, 274.274119348094, 
    275.280808254502, 276.286353887245, 277.287064312339, 278.294444793276, 
    279.296772014594, 270.333066283904, 271.33877455992, 272.345842319903, 
    273.350858180493, 274.353972278505, 275.360454510107, 276.365088896161, 
    277.369166956941, 278.372571708911, 279.38017503079), distanceToTx = c(80.255266401689, 
    80.156059067023, 79.98823695539, 79.826647129071, 79.76678667135, 
    79.788239825292, 79.734539327997, 79.74766421514, 79.801243848241, 
    79.765920888341, 80.255266401689, 80.15850240049, 79.98823695539, 
    79.826647129071, 79.76678667135, 79.788239825292, 79.735078924078, 
    79.74766421514, 79.801243848241, 79.764622734914, 80.251248121732, 
    80.146436869316, 79.984682320466, 79.82292012342, 79.761908518748, 
    79.796988776281, 79.736920997657, 79.745038376718, 79.802638836686, 
    79.770029970452, 80.243475525691, 80.127918207499, 79.978303140866, 
    79.816259117883, 79.749322030693, 79.809916018889, 79.744456560867, 
    79.738655068783, 79.788697533211, 79.784288359619, 80.260412958482, 
    80.168426829066, 79.992034911214, 79.830845773284, 79.7756751763, 
    79.778156038931, 79.732399593756, 79.752769548846, 79.799967731078, 
    79.757585110481, 80.251248121732, 80.146436869316, 79.984682320466, 
    79.822062073459, 79.75884601899, 79.801590491435, 79.738335109094, 
    79.74347007248, 79.803215965043, 79.771471198955, 80.250257298678, 
    80.146436869316, 79.983831684476, 79.822062073459, 79.75884601899, 
    79.801590491435, 79.738335109094, 79.74347007248, 79.803849157574, 
    79.771471198955, 80.243475525691, 80.130180105198, 79.978303140866, 
    79.816881283718, 79.749322030693, 79.80984572883, 79.744456560867, 
    79.738655068783, 79.790548644175, 79.784288359619, 80.246349000313, 
    80.137056554491, 79.980581246037, 79.818924707937, 79.753176142361, 
    79.808777040341, 79.741609845588, 79.740770913572, 79.796316397253, 
    79.777593733292, 80.238796415443, 80.119021911134, 79.974810568944, 
    79.814065350562, 79.743657315504, 79.810146783217, 79.749945098869, 
    79.737122584544, 79.781650522348, 79.791554933936), headerNoError = c(0.99999999989702, 
    0.9999999999981, 0.99999999999946, 0.9999999928026, 0.99999873265475, 
    0.77080141574964, 0.99007491438593, 0.99994396605059, 0.45588747062284, 
    0.93484381262491, 0.99999999989702, 0.99999999999816, 0.99999999999946, 
    0.9999999928026, 0.99999873265475, 0.77080141574964, 0.99008458785106, 
    0.99994396605059, 0.45588747062284, 0.93480223051707, 0.99999999989735, 
    0.99999999999789, 0.99999999999946, 0.99999999287551, 0.99999876302649, 
    0.46903147501117, 0.98835168988253, 0.99994427085086, 0.45235035271542, 
    0.93496741877335, 0.99999999989803, 0.99999999999781, 0.99999999999948, 
    0.99999999318224, 0.99994254156311, 0.46891362282273, 0.93382613917348, 
    0.99994594904099, 0.93002915596843, 0.93569767251247, 0.99999999989658, 
    0.99999999998074, 0.99999999999946, 0.99999999272802, 0.99999871586781, 
    0.76935240919896, 0.99002587758346, 0.99999881589732, 0.46179415706093, 
    0.93417422376389, 0.99999999989735, 0.99999999999789, 0.99999999999946, 
    0.99999999289347, 0.99999876940486, 0.46930769326427, 0.98837353639905, 
    0.99994447154714, 0.16313586712094, 0.93500824170148, 0.99999999989744, 
    0.99999999999789, 0.99999999999946, 0.99999999289347, 0.99999876940486, 
    0.46930769326427, 0.98837353639905, 0.99994447154714, 0.16330039178981, 
    0.93500824170148, 0.99999999989803, 0.99999999999781, 0.99999999999948, 
    0.99999999316541, 0.99994254156311, 0.46794586553266, 0.93382613917348, 
    0.99994594904099, 0.9303627789484, 0.93569767251247, 0.99999999989778, 
    0.9999999999978, 0.99999999999948, 0.99999999311433, 0.99999878195152, 
    0.47101897739483, 0.93368891853679, 0.99994556595217, 0.7571113417265, 
    0.93553999975802, 0.99999999998191, 0.99999999999784, 0.99999999999971, 
    0.99999891129658, 0.99994309267792, 0.46510628979591, 0.93442584181035, 
    0.99894450514543, 0.99890078483692, 0.76933812306423), receivedPower_dbm = c(-93.023492290586, 
    -92.388378035287, -92.205716340607, -93.816400586752, -95.023489422885, 
    -100.86308557253, -98.464763536915, -96.175707680373, -102.06189538385, 
    -99.716653422746, -93.023492290586, -92.384760627397, -92.205716340607, 
    -93.816400586752, -95.023489422885, -100.86308557253, -98.464201120719, 
    -96.175707680373, -102.06189538385, -99.717150021506, -93.022927803442, 
    -92.404017215549, -92.204561341714, -93.814319484729, -95.016990717792, 
    -102.01669022332, -98.558088145955, -96.173817001483, -102.07406915124, 
    -99.71517574876, -93.021813165972, -92.409586309743, -92.20229160243, 
    -93.805335867418, -96.184419849593, -102.01709540787, -99.728735187547, 
    -96.163233028048, -99.772547164798, -99.706399753853, -93.024204617071, 
    -92.745813384859, -92.206884754512, -93.818508150122, -95.027018807793, 
    -100.87000577258, -98.467607232407, -95.005311380324, -102.04157607608, 
    -99.724619517, -93.022927803442, -92.404017215549, -92.204561341714, 
    -93.813803344588, -95.015606885523, -102.0157405687, -98.556982278361, 
    -96.172566862738, -103.21871579865, -99.714687230796, -93.022787428238, 
    -92.404017215549, -92.204274688493, -93.813803344588, -95.015606885523, 
    -102.0157405687, -98.556982278361, -96.172566862738, -103.21784988098, 
    -99.714687230796, -93.021813165972, -92.409950613665, -92.20229160243, 
    -93.805838770576, -96.184419849593, -102.02042267497, -99.728735187547, 
    -96.163233028048, -99.768774335378, -99.706399753853, -93.022228914406, 
    -92.411048503835, -92.203136463155, -93.807357409082, -95.012865008237, 
    -102.00985717796, -99.730352912911, -96.165675535906, -100.92744056572, 
    -99.708301333236, -92.735781110993, -92.408137395049, -92.119533319039, 
    -94.982938427575, -96.181073124017, -102.03018610927, -99.721633629806, 
    -97.32940323644, -97.347613268692, -100.87007386786), snr = c(49.848348091678, 
    57.698190927109, 60.17669971462, 41.529809724535, 31.452202106925, 
    8.1976890851341, 14.240447804094, 24.122884195464, 6.2202875499406, 
    10.674183333671, 49.848348091678, 57.746270018264, 60.17669971462, 
    41.529809724535, 31.452202106925, 8.1976890851341, 14.242292077376, 
    24.122884195464, 6.2202875499406, 10.672962852322, 49.854827699773, 
    57.49079026127, 60.192705735317, 41.549715223147, 31.499301851462, 
    6.2853718719014, 13.937702343688, 24.133388256416, 6.2028757927148, 
    10.677815810561, 49.867624820879, 57.417115267867, 60.224172277442, 
    41.635752021705, 24.074540962859, 6.2847854917092, 10.644529778044, 
    24.19227425387, 10.537686730745, 10.699414795917, 49.84017267426, 
    53.139646558768, 60.160512118809, 41.509660845114, 31.42665220053, 
    8.1846370024428, 14.231126423354, 31.584125885363, 6.2494585568733, 
    10.654622041348, 49.854827699773, 57.49079026127, 60.192705735317, 
    41.55465351989, 31.509340361646, 6.2867464196657, 13.941251828322, 
    24.140336174865, 4.765718874642, 10.679016976694, 49.856439162736, 
    57.49079026127, 60.196678846453, 41.55465351989, 31.509340361646, 
    6.2867464196657, 13.941251828322, 24.140336174865, 4.7666691818074, 
    10.679016976694, 49.867624820879, 57.412299088098, 60.224172277442, 
    41.630930975211, 24.074540962859, 6.279972363168, 10.644529778044, 
    24.19227425387, 10.546845071479, 10.699414795917, 49.862851240855, 
    57.397787176282, 60.212457625018, 41.61637603957, 31.529239767749, 
    6.2952688513108, 10.640565481982, 24.178672145334, 8.0771089950663, 
    10.694731030907, 53.262541905639, 57.43627424514, 61.382796189332, 
    31.747253311549, 24.093100244121, 6.2658701281075, 10.661949889074, 
    18.495227442305, 18.417839037171, 8.1845086722809), frameId = c(15051, 
    15106, 15165, 15220, 15279, 15330, 15385, 15452, 15511, 15566, 
    15019, 15074, 15129, 15184, 15239, 15298, 15353, 15412, 15471, 
    15526, 14947, 14994, 15057, 15112, 15171, 15226, 15281, 15332, 
    15391, 15442, 14971, 15030, 15085, 15144, 15203, 15262, 15321, 
    15380, 15435, 15490, 14915, 14978, 15033, 15092, 15147, 15198, 
    15257, 15312, 15371, 15430, 14975, 15034, 15089, 15140, 15195, 
    15254, 15313, 15368, 15427, 15478, 14987, 15046, 15105, 15160, 
    15215, 15274, 15329, 15384, 15447, 15506, 14943, 15002, 15061, 
    15116, 15171, 15230, 15285, 15344, 15399, 15454, 14971, 15026, 
    15081, 15136, 15195, 15258, 15313, 15368, 15423, 15478, 15039, 
    15094, 15149, 15204, 15263, 15314, 15369, 15428, 15487, 15546
    ), packetOkSinr = c(0.99999999314881, 0.9999999998736, 0.99999999996428, 
    0.99999952114066, 0.99991568416005, 3.00628034688444e-08, 
    0.51497487795954, 0.99627877136019, 0, 0.011303253101957, 
    0.99999999314881, 0.99999999987726, 0.99999999996428, 0.99999952114066, 
    0.99991568416005, 3.00628034688444e-08, 0.51530974419663, 
    0.99627877136019, 0, 0.011269851265775, 0.9999999931708, 
    0.99999999985986, 0.99999999996428, 0.99999952599145, 0.99991770469509, 
    0, 0.45861812482641, 0.99629897628155, 0, 0.011403119534097, 
    0.99999999321568, 0.99999999985437, 0.99999999996519, 0.99999954639936, 
    0.99618434878558, 0, 0.010513119213425, 0.99641022914441, 
    0.00801687746446111, 0.012011103529927, 0.9999999931195, 
    0.99999999871861, 0.99999999996428, 0.99999951617905, 0.99991456738049, 
    2.6525298291169e-08, 0.51328066587104, 0.9999212220316, 0, 
    0.010777054258914, 0.9999999931708, 0.99999999985986, 0.99999999996428, 
    0.99999952718674, 0.99991812902805, 0, 0.45929307038653, 
    0.99631228046814, 0, 0.011436292559188, 0.99999999317629, 
    0.99999999985986, 0.99999999996428, 0.99999952718674, 0.99991812902805, 
    0, 0.45929307038653, 0.99631228046814, 0, 0.011436292559188, 
    0.99999999321568, 0.99999999985437, 0.99999999996519, 0.99999954527918, 
    0.99618434878558, 0, 0.010513119213425, 0.99641022914441, 
    0.00821047996950475, 0.012011103529927, 0.99999999319919, 
    0.99999999985345, 0.99999999996519, 0.99999954188106, 0.99991896371849, 
    0, 0.010410830482692, 0.996384831822, 9.12484388049251e-09, 
    0.011877185067536, 0.99999999879646, 0.9999999998562, 0.99999999998077, 
    0.99992756868677, 0.9962208785486, 0, 0.010971897073662, 
    0.93214999078663, 0.92943956665979, 2.64925478221656e-08), 
    snir = c(49.848348091678, 57.698190927109, 60.17669971462, 
    41.529809724535, 31.452202106925, 8.1976890851341, 14.240447804094, 
    24.122884195464, 6.2202875499406, 10.674183333671, 49.848348091678, 
    57.746270018264, 60.17669971462, 41.529809724535, 31.452202106925, 
    8.1976890851341, 14.242292077376, 24.122884195464, 6.2202875499406, 
    10.672962852322, 49.854827699773, 57.49079026127, 60.192705735317, 
    41.549715223147, 31.499301851462, 6.2853718719014, 13.937702343688, 
    24.133388256416, 6.2028757927148, 10.677815810561, 49.867624820879, 
    57.417115267867, 60.224172277442, 41.635752021705, 24.074540962859, 
    6.2847854917092, 10.644529778044, 24.19227425387, 10.537686730745, 
    10.699414795917, 49.84017267426, 53.139646558768, 60.160512118809, 
    41.509660845114, 31.42665220053, 8.1846370024428, 14.231126423354, 
    31.584125885363, 6.2494585568733, 10.654622041348, 49.854827699773, 
    57.49079026127, 60.192705735317, 41.55465351989, 31.509340361646, 
    6.2867464196657, 13.941251828322, 24.140336174865, 4.765718874642, 
    10.679016976694, 49.856439162736, 57.49079026127, 60.196678846453, 
    41.55465351989, 31.509340361646, 6.2867464196657, 13.941251828322, 
    24.140336174865, 4.7666691818074, 10.679016976694, 49.867624820879, 
    57.412299088098, 60.224172277442, 41.630930975211, 24.074540962859, 
    6.279972363168, 10.644529778044, 24.19227425387, 10.546845071479, 
    10.699414795917, 49.862851240855, 57.397787176282, 60.212457625018, 
    41.61637603957, 31.529239767749, 6.2952688513108, 10.640565481982, 
    24.178672145334, 8.0771089950663, 10.694731030907, 53.262541905639, 
    57.43627424514, 61.382796189332, 31.747253311549, 24.093100244121, 
    6.2658701281075, 10.661949889074, 18.495227442305, 18.417839037171, 
    8.1845086722809), ookSnirBer = c(8.8808636558081e-24, 3.2219795637026e-27, 
    2.6468895519653e-28, 3.9807779074715e-20, 1.0849324265615e-15, 
    2.5705217057696e-05, 4.7313805615763e-08, 1.8800438086075e-12, 
    0.00021005320203921, 1.9147343768384e-06, 8.8808636558081e-24, 
    3.0694773489537e-27, 2.6468895519653e-28, 3.9807779074715e-20, 
    1.0849324265615e-15, 2.5705217057696e-05, 4.7223753038869e-08, 
    1.8800438086075e-12, 0.00021005320203921, 1.9171738578051e-06, 
    8.8229427230445e-24, 3.9715925056443e-27, 2.6045198111088e-28, 
    3.9014083702734e-20, 1.0342658440386e-15, 0.00019591630514278, 
    6.4692014108683e-08, 1.8600094209271e-12, 0.0002140067535655, 
    1.9074922485477e-06, 8.7096574467175e-24, 4.2779443633862e-27, 
    2.5231916788231e-28, 3.5761615214425e-20, 1.9750692814982e-12, 
    0.0001960392878411, 1.9748966344895e-06, 1.7515881895994e-12, 
    2.2078334799411e-06, 1.8649940680806e-06, 8.954486301678e-24, 
    3.2021085732779e-25, 2.690441113724e-28, 4.0627628846548e-20, 
    1.1134484878561e-15, 2.6061691733331e-05, 4.777159157954e-08, 
    9.4891388749738e-16, 0.00020359398491544, 1.9542110660398e-06, 
    8.8229427230445e-24, 3.9715925056443e-27, 2.6045198111088e-28, 
    3.8819641115984e-20, 1.0237769828158e-15, 0.00019562832342849, 
    6.4455095380046e-08, 1.8468752030971e-12, 0.0010099091367628, 
    1.9051035165106e-06, 8.8085966897635e-24, 3.9715925056443e-27, 
    2.594108048185e-28, 3.8819641115984e-20, 1.0237769828158e-15, 
    0.00019562832342849, 6.4455095380046e-08, 1.8468752030971e-12, 
    0.0010088638355194, 1.9051035165106e-06, 8.7096574467175e-24, 
    4.2987746909572e-27, 2.5231916788231e-28, 3.593647329558e-20, 
    1.9750692814982e-12, 0.00019705170257492, 1.9748966344895e-06, 
    1.7515881895994e-12, 2.1868296425817e-06, 1.8649940680806e-06, 
    8.7517439682173e-24, 4.3621551072316e-27, 2.553168170837e-28, 
    3.6469582463164e-20, 1.0032983660212e-15, 0.00019385229409318, 
    1.9830820164805e-06, 1.7760568361323e-12, 2.919419915209e-05, 
    1.8741284335866e-06, 2.8285944348148e-25, 4.1960751547207e-27, 
    7.8468215407139e-29, 8.0407329049747e-16, 1.9380328071065e-12, 
    0.00020004849911333, 1.9393279417733e-06, 5.9354475879597e-10, 
    6.4258355913627e-10, 2.6065221215415e-05), ookSnrBer = c(8.8808636558081e-24, 
    3.2219795637026e-27, 2.6468895519653e-28, 3.9807779074715e-20, 
    1.0849324265615e-15, 2.5705217057696e-05, 4.7313805615763e-08, 
    1.8800438086075e-12, 0.00021005320203921, 1.9147343768384e-06, 
    8.8808636558081e-24, 3.0694773489537e-27, 2.6468895519653e-28, 
    3.9807779074715e-20, 1.0849324265615e-15, 2.5705217057696e-05, 
    4.7223753038869e-08, 1.8800438086075e-12, 0.00021005320203921, 
    1.9171738578051e-06, 8.8229427230445e-24, 3.9715925056443e-27, 
    2.6045198111088e-28, 3.9014083702734e-20, 1.0342658440386e-15, 
    0.00019591630514278, 6.4692014108683e-08, 1.8600094209271e-12, 
    0.0002140067535655, 1.9074922485477e-06, 8.7096574467175e-24, 
    4.2779443633862e-27, 2.5231916788231e-28, 3.5761615214425e-20, 
    1.9750692814982e-12, 0.0001960392878411, 1.9748966344895e-06, 
    1.7515881895994e-12, 2.2078334799411e-06, 1.8649940680806e-06, 
    8.954486301678e-24, 3.2021085732779e-25, 2.690441113724e-28, 
    4.0627628846548e-20, 1.1134484878561e-15, 2.6061691733331e-05, 
    4.777159157954e-08, 9.4891388749738e-16, 0.00020359398491544, 
    1.9542110660398e-06, 8.8229427230445e-24, 3.9715925056443e-27, 
    2.6045198111088e-28, 3.8819641115984e-20, 1.0237769828158e-15, 
    0.00019562832342849, 6.4455095380046e-08, 1.8468752030971e-12, 
    0.0010099091367628, 1.9051035165106e-06, 8.8085966897635e-24, 
    3.9715925056443e-27, 2.594108048185e-28, 3.8819641115984e-20, 
    1.0237769828158e-15, 0.00019562832342849, 6.4455095380046e-08, 
    1.8468752030971e-12, 0.0010088638355194, 1.9051035165106e-06, 
    8.7096574467175e-24, 4.2987746909572e-27, 2.5231916788231e-28, 
    3.593647329558e-20, 1.9750692814982e-12, 0.00019705170257492, 
    1.9748966344895e-06, 1.7515881895994e-12, 2.1868296425817e-06, 
    1.8649940680806e-06, 8.7517439682173e-24, 4.3621551072316e-27, 
    2.553168170837e-28, 3.6469582463164e-20, 1.0032983660212e-15, 
    0.00019385229409318, 1.9830820164805e-06, 1.7760568361323e-12, 
    2.919419915209e-05, 1.8741284335866e-06, 2.8285944348148e-25, 
    4.1960751547207e-27, 7.8468215407139e-29, 8.0407329049747e-16, 
    1.9380328071065e-12, 0.00020004849911333, 1.9393279417733e-06, 
    5.9354475879597e-10, 6.4258355913627e-10, 2.6065221215415e-05
    )), class = "data.frame", row.names = c(NA, -100L), .Names = c("run", 
"repetition", "module", "configname", "packetByteLength", "numVehicles", 
"dDistance", "time", "distanceToTx", "headerNoError", "receivedPower_dbm", 
"snr", "frameId", "packetOkSinr", "snir", "ookSnirBer", "ookSnrBer"
))

Finding the transformation function

  1. y1 --> y2 This function is used to transform the data of the secondary y axis to be "normalized" according to the first y axis

transformation function: f(y1) = 0.025*x + 2.75

  1. y2 --> y1 This function is used to transform the break points of the first y axis to the values of the second y axis. Note that the axis are swapped now.

transformation function: f(y1) = 40*x - 110

Plotting

Note how the transformation functions are used in the ggplot call to transform the data "on-the-fly"

ggplot(data=combined_80_8192 %>% filter (time > 270, time < 280), aes(x=time) ) +
  stat_summary(aes(y=receivedPower_dbm ), fun.y=mean, geom="line", colour="black") +
  stat_summary(aes(y=packetOkSinr*40 - 110 ), fun.y=mean, geom="line", colour="black", position = position_dodge(width=10)) +
  scale_x_continuous() +
  scale_y_continuous(breaks = seq(-0,-110,-10), "y_first", sec.axis=sec_axis(~.*0.025+2.75, name="y_second") ) 

The first stat_summary call is the one that sets the base for the first y axis. The second stat_summary call is called to transform the data. Remember that all of the data will take as base the first y axis. So that data needs to be normalized for the first y axis. To do that I use the transformation function on the data: y=packetOkSinr*40 - 110

Now to transform the second axis I use the opposite function within the scale_y_continuous call: sec.axis=sec_axis(~.*0.025+2.75, name="y_second") .


I acknowledge and agree with hadley (and others), that separate y-scales are "fundamentally flawed". Having said that – I often wish ggplot2 had the feature – particularly, when the data is in wide-format and I quickly want to visualise or check the data (ie for personal use only).

While the tidyverse library makes it fairly easy to convert the data to long-format (such that facet_grid() will work), the process is still not trivial, as seen below:

library(tidyverse)
df.wide %>%
    # Select only the columns you need for the plot.
    select(date, column1, column2, column3) %>%
    # Create an id column – needed in the `gather()` function.
    mutate(id = n()) %>%
    # The `gather()` function converts to long-format. 
    # In which the `type` column will contain three factors (column1, column2, column3),
    # and the `value` column will contain the respective values.
    # All the while we retain the `id` and `date` columns.
    gather(type, value, -id, -date) %>%
    # Create the plot according to your specifications
    ggplot(aes(x = date, y = value)) +
        geom_line() +
        # Create a panel for each `type` (ie. column1, column2, column3).
        # If the types have different scales, you can use the `scales="free"` option.
        facet_grid(type~., scales = "free")

Taking above answers and some fine-tuning (and for whatever it's worth), here is a way of achieving two scales via sec_axis :

Assume a simple (and purely fictional) data set dt : for five days, it tracks the number of interruptions VS productivity:

        when numinter prod
1 2018-03-20        1 0.95
2 2018-03-21        5 0.50
3 2018-03-23        4 0.70
4 2018-03-24        3 0.75
5 2018-03-25        4 0.60

(the ranges of both columns differ by about factor 5).

The following code will draw both series that they use up the whole y axis:

ggplot() + 
  geom_bar(mapping = aes(x = dt$when, y = dt$numinter), stat = "identity", fill = "grey") +
  geom_line(mapping = aes(x = dt$when, y = dt$prod*5), size = 2, color = "blue") + 
  scale_x_date(name = "Day", labels = NULL) +
  scale_y_continuous(name = "Interruptions/day", 
    sec.axis = sec_axis(~./5, name = "Productivity % of best", 
      labels = function(b) { paste0(round(b * 100, 0), "%")}))

Here's the result (above code + some color tweaking):

The point (aside from using sec_axis when specifying the y_scale is to multiply each value the 2nd data series with 5 when specifying the series. In order to get the labels right in the sec_axis definition, it then needs dividing by 5 (and formatting). So a crucial part in above code is really *5 in the geom_line and ~./5 in sec_axis (a formula dividing the current value . by 5).

In comparison (I don't want to judge the approaches here), this is how two charts on top of one another look like:

You can judge for yourself which one better transports the message (“Don't disrupt people at work!”). Guess that's a fair way to decide.

Der vollständige Code für beide Bilder (es ist nicht wirklich mehr als das, was oben ist, nur komplett und bereit zu laufen) ist hier: https://gist.github.com/sebastianrothbucher/de847063f32fdff02c83b75f59c36a7d eine genauere Erklärung hier: https://sebastianrothbucher.github.io/datascience/r/visualization/ggplot/2018/03/24/two-scales-ggplot-r.html


We definitely could build a plot with dual Y-axises using base R funtion plot .

# pseudo dataset
df <- data.frame(x = seq(1, 1000, 1), y1 = sample.int(100, 1000, replace=T), y2 = sample(50, 1000, replace = T))

# plot first plot 
with(df, plot(y1 ~ x, col = "red"))

# set new plot
par(new = T) 

# plot second plot, but without axis
with(df, plot(y2 ~ x, type = "l", xaxt = "n", yaxt = "n", xlab = "", ylab = ""))

# define y-axis and put y-labs
axis(4)
with(df, mtext("y2", side = 4))

You can use facet_wrap(~ variable, ncol= ) on a variable to create a new comparison. It's not on the same axis, but it is similar.





r-faq