public void Test_LinearRegression_RandomXs()
{
int pointCount = 50;
double actualSlope = 3;
double actualOffset = 200;
(double[] xs, double[] ys) = GetNoisyLinearData_RandomlySpaced(pointCount, actualSlope, actualOffset);
// fit the random data with the linear regression model
var model = new ScottPlot.Statistics.LinearRegressionLine(xs, ys);
// plot to visually assess goodness of fit
var plt = new ScottPlot.Plot(450, 300);
plt.Title($"Y = {model.slope:0.0000}x + {model.offset:0.0}\nR² = {model.rSquared:0.0000}");
plt.AddScatterPoints(xs, ys);
// ADD THIS BACK IN!
//plt.PlotLine(model.slope, model.offset, (xs.Min(), xs.Max()), lineWidth: 2, label: "model", lineStyle: ScottPlot.LineStyle.Dash);
//plt.PlotLine(actualSlope, actualOffset, (xs.Min(), xs.Max()), lineWidth: 2, label: "actual");
plt.Legend();
TestTools.SaveFig(plt);
// ensure the fit is good
Assert.AreEqual(actualSlope, model.slope, .5);
Assert.AreEqual(actualOffset, model.offset, 10);
}
public void GetCoefficients_FirstXNotZero_ResultEqualtoTI_84_Plus()
{
double[] x = new double[] { 1, 2, 3, 4, 5, 6, 7 };
double[] y = new double[] { 2, 2, 3, 3, 3.8, 4.2, 4 };
var reg = new ScottPlot.Statistics.LinearRegressionLine(x, y);
Assert.AreEqual(0.4, reg.slope, 0.0001);
Assert.AreEqual(1.5428, reg.offset, 0.0001);
}
public void ExecuteRecipe(Plot plt)
{
// Create some linear but noisy data
double[] ys = DataGen.NoisyLinear(null, pointCount: 100, noise: 30);
double[] xs = DataGen.Consecutive(ys.Length);
double x1 = xs[0];
double x2 = xs[xs.Length - 1];
// use the linear regression fitter to fit these data
var model = new ScottPlot.Statistics.LinearRegressionLine(xs, ys);
// plot the original data and add the regression line
plt.Title($"Y = {model.slope:0.0000}x + {model.offset:0.0}\nR² = {model.rSquared:0.0000}");
plt.AddScatter(xs, ys, lineWidth: 0);
plt.AddLine(model.slope, model.offset, (x1, x2), lineWidth: 2);
}
private void CreateScatterAndFitRegression()
{
Random rand = new Random();
// generate linear data with noise
double slope = rand.Next(-1000, 1000) / 100.0;
double offset = rand.Next(-1000, 1000);
int pointCount = rand.Next(50, 1000);
double[] ys = ScottPlot.DataGen.NoisyLinear(
rand: rand,
pointCount: pointCount,
slope: slope,
offset: offset,
noise: rand.Next(10, 500)
);
// create matching X-axis ticks
double pointSpacing = rand.Next(1, 1000) / 1000.0;
double[] xs = ScottPlot.DataGen.Consecutive(ys.Length, spacing: pointSpacing);
realLineLabel.Text = string.Format("Y = {0:0.00}x + {1:0.00} (n={2})", slope / pointSpacing, offset, pointCount);
// plot data
formsPlot1.plt.Clear();
formsPlot1.plt.PlotScatter(xs, ys, lineWidth: 0);
formsPlot1.plt.AxisAuto();
// perform the linear regression
var linreg = new ScottPlot.Statistics.LinearRegressionLine(xs, ys);
double fittedSlope = linreg.slope;
fittedLineLabel.Text = string.Format("Y = {0:0.00}x + {1:0.00}", fittedSlope, linreg.offset);
// plot the linear regression line (as a scatter plot with just two points)
double x1 = xs.First();
double x2 = xs.Last();
double y1 = linreg.GetValueAt(xs.First());
double y2 = linreg.GetValueAt(xs.Last());
formsPlot1.plt.PlotLine(x1, y1, x2, y2, lineWidth: 3, color: Color.Black);
// force a redraw of the user control
formsPlot1.Render();
}
public void Test_LinearRegression_EvenlySpacedXs()
{
Random rand = new Random(0);
for (int i = 0; i < 1000; i++)
{
// create random data
(double slope, double offset, double[] xs, double[] data) = GetRandomRegressionData(rand);
// fit the random data with the linear regression model
var model = new ScottPlot.Statistics.LinearRegressionLine(data, 0, 1);
// ensure the model's coeffecients are similar to the real ones
double slopeError = Math.Abs(slope - model.slope);
double offsetError = Math.Abs(offset - model.offset);
Assert.That(slopeError, Is.LessThan(.1));
Assert.That(offsetError, Is.LessThan(10));
}
}
public void Render(Plot plt)
{
// Create some linear but noisy data
Random rand = new Random(0);
double[] ys = ScottPlot.DataGen.NoisyLinear(rand, pointCount: 100, noise: 30);
double[] xs = ScottPlot.DataGen.Consecutive(ys.Length);
double x1 = xs[0];
double x2 = xs[xs.Length - 1];
// use the linear regression fitter to fit these data
var model = new ScottPlot.Statistics.LinearRegressionLine(xs, ys);
double y1 = model.GetValueAt(x1);
double y2 = model.GetValueAt(x2);
// plot the original data and add the regression line
plt.PlotScatter(xs, ys, lineWidth: 0, label: "original data");
plt.PlotLine(x1, y1, x2, y2, lineWidth: 3, label: "linear regression");
plt.Legend();
plt.Title(model.ToString());
}
private void formsPlot1_Load(object sender, EventArgs e)
{
// this is the plotting place - all in the Load
// string tname1 = "WXROVER";
// string tname2 = "WXIN";
// int pointcount = 5000;
// double[,] tempData;
int cnt = 0;
// double[] t1 = new double[Globals.pointcount];
// double[] t2 = new double[Globals.pointcount];
// a couple of local lists
List <double> orgs = new List <double> {
};
List <double> chks = new List <double> {
};
classWXstnChk WXchk = new classWXstnChk(); //make an object WXck
WXchk.cnxstring("192.168.1.15", "DAWES_SQL2008", "WeatherStation", "WeatherStation", "Esp32a.b."); //a constructor - no arguments
//go get the data
Globals.tempData = WXchk.getTemperatureData(Globals.pointcount, Globals.tname1, Globals.tname2); //this gets the data as an XY pair
//there will be a lot of duplcates in this array. Lets remove the dupes - done in the SQL
//for scotterplot to work, these need to be in two separate Arrays, so we seperate them out in
//this loop and we might as well get rid of the empty slots
//maybe this would also work
// double [] tab1 = new double[tempData.Length];
for (int i = 0; i < pointcount - 1; i++)
{
if (Globals.tempData[i, 0] > 0)
{
t1[i] = tempData[i, 0];
t2[i] = tempData[i, 1];
t1qq[i] = tempData[i, 0];
t2qq[i] = tempData[i, 1];
//might as well put the data in the grid
dgXY.Rows.Add();
try
{
dgXY.Rows[i].Cells["dgXYrecord"].Value = i.ToString(); //puts the record no on the dgXY
dgXY.Rows[i].Cells["dgXYx"].Value = t1[i];
dgXY.Rows[i].Cells["dgXYy"].Value = t2[i];
}
catch { }
cnt += 1;
}
}//end of for loop
// we have two arrays lets sort them to form a QQ plot
Array.Sort(t1qq);
Array.Sort(t2qq);
// the arrays are declared a certain size and are now full of 0's
//this will reduce them down to just data.
// if there is no data in the arrays then cnt will = 0
if (cnt == 0)
{
MessageBox.Show("No Data in your selection.", "No data", MessageBoxButtons.OK, MessageBoxIcon.Exclamation);
return;
}
Array.Resize(ref t1, cnt);
Array.Resize(ref t2, cnt);
cntG = cnt;
// int aa = cnt;
int ab = t2.GetLength(0);
// create a Population object from the data
var popt1 = new ScottPlot.Statistics.Population(t1);
var popt2 = new ScottPlot.Statistics.Population(t2);
// now lets fill the stats box
lblT1.Text = tname1;
lblT2.Text = tname2;
lblMeanT1.Text = Math.Round(popt1.mean, 2).ToString();
lblMeanT2.Text = Math.Round(popt2.mean, 2).ToString();
lblMedT1.Text = Math.Round(popt1.median, 2).ToString();
lblMedT2.Text = Math.Round(popt2.median, 2).ToString();
lblQ3T1.Text = popt1.Q3.ToString();
lblq3table2.Text = popt1.Q3.ToString();
lblQ1t1.Text = popt1.Q1.ToString();
lblQ1T2.Text = popt2.Q1.ToString();
lblstderrT1.Text = Math.Round(popt1.stdErr, 2).ToString();
lblstderrT2.Text = Math.Round(popt2.stdErr, 2).ToString();
lblstdevT1.Text = Math.Round(popt1.stDev, 2).ToString();
lblstdevT2.Text = Math.Round(popt2.stDev, 2).ToString();
lblIQRt1.Text = Math.Round(popt1.IQR, 2).ToString();
lblIQRt2.Text = Math.Round(popt2.IQR, 2).ToString();
lblNtn1.Text = cnt.ToString(); //opt1.aa.ToString();
// lblNtn2.Text = ab.ToString(); // popt2.count.ToString();
//this gives us the regression line and the correlation coeficient
var model = new ScottPlot.Statistics.LinearRegressionLine(t1, t2);
string line = $"Y={model.slope:0.0000}X + {model.offset:0.0}";
lblLine.Text = line;
lblCorrel.Text = $"R² = {model.rSquared:0.0000}";
formsPlot1.plt.Title($"ScarpWeather Temperature XY Plot");
formsPlot1.plt.XLabel(tname1);
formsPlot1.plt.YLabel(tname2);
//formsPlot1.plt.AxisAuto(horizontalMargin: 0, verticalMargin: 0.5);
// xmin,xmax,ymin,ymax
int axisMin, axisMax;
axisMin = (int)popt1.min - 2; //get some data from the arrays
//axisMin = 0;
axisMax = (int)popt1.max + 2;
formsPlot1.plt.Axis(axisMin, axisMax, axisMin, axisMax); //defines the min and max of the chart
// formsPlot1.plt.Axis(15,40,15,40);
formsPlot1.plt.Legend(); // got ot have this be we toggle it with hte context menu
//this draws the scatter plot
XYplotPoints = formsPlot1.plt.PlotScatter(t1, t2, Color.DarkCyan, lineWidth: 0, markerSize: 5, label: "XYData");
//draw the QQ plot
QQplotPoints = formsPlot1.plt.PlotScatter(t1qq, t2qq, Color.DarkGreen, lineWidth: 0, markerSize: 5, label: "QQData");
QQplotPoints.visible = false;
// draws the regresion line
//if (plotRegression)
regLine = formsPlot1.plt.PlotLine(model.slope, model.offset, (axisMin, axisMax), lineWidth: 2, label: "Regression", color: Color.BlueViolet);
//plot a 45 degree line = X=Y
XequalsYLine = formsPlot1.plt.PlotLine(1.0, 0.0, (0, 100), Color.Goldenrod, lineWidth: 2, label: "X=Y", lineStyle: ScottPlot.LineStyle.Dot); //uses the equation of the line to plot
//the 10% error lines
errorLineA = formsPlot1.plt.PlotLine(0, 0, 100.0 - 0.1 * 100, 100, color: Color.OrangeRed, lineWidth: 2, label: "10%error", lineStyle: ScottPlot.LineStyle.Dash);
errorLineB = formsPlot1.plt.PlotLine(0, 0, 100, 100.0 - 0.1 * 100, color: Color.OrangeRed, lineWidth: 2, lineStyle: ScottPlot.LineStyle.Dash);
formsPlot1.Render();//draw the chart on the PC screen
}//end function
public void RenderPlot()
{
plotFrame.plt.Clear();
RefreshTitleAndAxis(false);
bool containsDateAxis = false;
foreach (PlotParameters curr in ((App)App.Current).GetSeries())
{
if (curr.drawSettings.label == "")
{
curr.drawSettings.label = null;//Prevents it from showing up in the legend
}
if (curr.drawSettings.dateXAxis)
{
containsDateAxis = true;
object timeUnit;
object numTimeUnitsObj;
curr.metaData.TryGetValue("timeUnit", out timeUnit);
curr.metaData.TryGetValue("numTimeUnits", out numTimeUnitsObj);
ScottPlot.Config.DateTimeUnit dateTimeUnit = (ScottPlot.Config.DateTimeUnit)timeUnit;
int numTimeUnits = (int)numTimeUnitsObj;
dateUnit = dateTimeUnit;
dateUnitSpacing = numTimeUnits;
if (dateTimeUnit == ScottPlot.Config.DateTimeUnit.Year)//Grid spacing of one year is currently unsupported -_-
{
plotFrame.plt.Grid(xSpacing: 12, xSpacingDateTimeUnit: ScottPlot.Config.DateTimeUnit.Month);
}
else
{
plotFrame.plt.Grid(xSpacing: numTimeUnits, xSpacingDateTimeUnit: dateTimeUnit);
}
plotFrame.plt.Ticks(dateTimeX: true, xTickRotation: 30, fontSize: 10); // Default fontSize is 12
}
switch (curr.drawSettings.type)
{
case PlotType.scatter:
double[] xsScatter = ((double[][])curr.data)[0];
double[] ysScatter = ((double[][])curr.data)[1];
if (logAxis)
{
ysScatter = ScottPlot.Tools.Log10(ysScatter);
}
if (!curr.hasErrorData)
{
plotFrame.plt.PlotScatterHighlight(xsScatter, ysScatter, curr.drawSettings.colour, curr.drawSettings.drawLine ? 1 : 0, label: curr.drawSettings.label, markerShape: curr.drawSettings.markerShape, highlightedColor: curr.drawSettings.colour);
}
else
{
double[] errorX = ((double[][])curr.errorData)[0];
double[] errorY = ((double[][])curr.errorData)[1];
plotFrame.plt.PlotScatterHighlight(xsScatter, ysScatter, curr.drawSettings.colour, curr.drawSettings.drawLine ? 1 : 0, label: curr.drawSettings.label, markerShape: curr.drawSettings.markerShape, errorX: errorX, errorY: errorY, highlightedColor: curr.drawSettings.colour);
}
if (curr.drawSettings.drawLinearRegression)
{
var model = new ScottPlot.Statistics.LinearRegressionLine(xsScatter, ysScatter);
double x1 = xsScatter[0];
double x2 = xsScatter[xsScatter.Length - 1];
double y1 = model.GetValueAt(x1);
double y2 = model.GetValueAt(x2);
plotFrame.plt.PlotLine(x1, y1, x2, y2, lineWidth: 3, label: $"ŷ = {model.offset:f9} + {model.slope:f9}x");
}
break;
case PlotType.signal:
object sampleRate = 100;
curr.metaData.TryGetValue("sampleRate", out sampleRate);
object xOffset = 0;
curr.metaData.TryGetValue("xOffset", out xOffset);
double[] data = (double[])curr.data;
if (logAxis)
{
data = ScottPlot.Tools.Log10(data);
}
plotFrame.plt.PlotSignalConst(data, (double)sampleRate, (double)xOffset, color: curr.drawSettings.colour, lineWidth: curr.drawSettings.drawLine ? 1 : 0, label: curr.drawSettings.label, markerSize: 0);
break;
case PlotType.bar:
double[] xsBar = ((double[][])curr.data)[0];
double[] ysBar = ((double[][])curr.data)[1];
if (logAxis)
{
ysBar = ScottPlot.Tools.Log10(ysBar);
}
if (!curr.hasErrorData)
{
plotFrame.plt.PlotBar(xsBar, ysBar, fillColor: curr.drawSettings.colour, outlineColor: curr.drawSettings.colour, errorColor: curr.drawSettings.colour, label: curr.drawSettings.label);
}
else
{
double[] errorY = ((double[])curr.errorData);
plotFrame.plt.PlotBar(xsBar, ysBar, fillColor: curr.drawSettings.colour, outlineColor: curr.drawSettings.colour, errorColor: curr.drawSettings.colour, label: curr.drawSettings.label, errorY: errorY);
}
break;
case PlotType.histogram:
ScottPlot.Statistics.Histogram histogram = new ScottPlot.Statistics.Histogram((double[])curr.data);
double[] yData = histogram.counts;
switch (curr.drawSettings.histogramType)
{
case HistogramType.fraction | HistogramType.density:
yData = histogram.countsFrac;
break;
case HistogramType.fraction | HistogramType.cumulative:
yData = histogram.cumulativeFrac;
break;
case HistogramType.count | HistogramType.cumulative:
yData = histogram.cumulativeCounts;
break;
}
plotFrame.plt.PlotBar(histogram.bins, yData, fillColor: curr.drawSettings.colour, outlineColor: curr.drawSettings.colour, errorColor: curr.drawSettings.colour, label: curr.drawSettings.label);
break;
case PlotType.horizontal_line:
double hLineData = (double)curr.data;
plotFrame.plt.PlotHLine(hLineData, color: curr.drawSettings.colour, label: curr.drawSettings.label);
break;
case PlotType.vertical_line:
double vLineData = (double)curr.data;
plotFrame.plt.PlotVLine(vLineData, color: curr.drawSettings.colour, label: curr.drawSettings.label);
break;
case PlotType.horizontal_span:
(double hSpanMin, double hSpanMax) = (ValueTuple <double, double>)curr.data;
plotFrame.plt.PlotHSpan(hSpanMin, hSpanMax, color: curr.drawSettings.colour, label: curr.drawSettings.label);
break;
case PlotType.vertical_span:
(double vSpanMin, double vSpanMax) = (ValueTuple <double, double>)curr.data;
plotFrame.plt.PlotVSpan(vSpanMin, vSpanMax, color: curr.drawSettings.colour, label: curr.drawSettings.label);
break;
case PlotType.box_whisker:
var plotObj = plotFrame.plt.PlotPopulations((ScottPlot.Statistics.Population)curr.data, label: curr.drawSettings.label);
plotObj.boxStyle = PlottablePopulations.BoxStyle.BoxMedianQuartileOutlier;
plotObj.displayDistributionCurve = false;
break;
case PlotType.function:
var f = (Func <double, double?>)curr.data;
plotFrame.plt.PlotFunction(f, label: curr.drawSettings.label, markerShape: MarkerShape.none, lineStyle: LineStyle.Dash);
break;
case PlotType.bar_grouped:
var plotData = ((double[][])curr.data);
string[] groupLabels = (string[])curr.metaData.GetValueOrDefault("group_names");
string[] seriesLabels = (string[])curr.metaData.GetValueOrDefault("series_names");
if (!curr.hasErrorData)
{
plotFrame.plt.PlotBarGroups(groupLabels, seriesLabels, plotData);
}
else
{
plotFrame.plt.PlotBarGroups(groupLabels, seriesLabels, plotData, (double[][])curr.errorData);
}
break;
}
}
isDateAxis = containsDateAxis;
plotFrame.plt.Legend();
(double highlightX, double highlightY) = (plotFrame.plt.Axis()[0], plotFrame.plt.Axis()[3]);
foreach (var curr in rawPlottables)
{
if (curr != snappedCoordinates || showCoordinates)
{
plotFrame.plt.Add(curr);//Got axed when we cleared everything
}
}
RenderFrame();
}
