pascal/lazarus-ccr
1
0
Fork 0
Code Issues Releases Activity

LazStats: More refactoring of BlkAnovaUnit (and OneCaseAnovaUnit)

Browse Source 
git-svn-id: https://svn.code.sf.net/p/lazarus-ccr/svn@7489 8e941d3f-bd1b-0410-a28a-d453659cc2b4
This commit is contained in:
wp_xxyyzz
2020-06-23 18:16:25 +00:00
parent 6a36491ee5
commit 66a428dbb0
3 changed files with 246 additions and 235 deletions
Download Patch File Download Diff File Expand all files Collapse all files

394
applications/lazstats/source/forms/analysis/comparisons/blkanovaunit.pas
View File

@ -86,7 +86,7 @@ type
private
{ private declarations }
FAutoSized: Boolean;
NoSelected, intvalue, N : integer;
NoSelected, N : integer;
ColNoSelected : IntDyneVec;
outline, cellstring : string;
SSDep, SSErr, SSF1, SSF2, SSF3, SSF1F2, SSF1F3, SSF2F3, SSF1F2F3 : double;
@ -99,19 +99,19 @@ type
DepVarCol, F1Col, F2Col, F3Col, Nf1cells, Nf2cells, Nf3cells : integer;
MeanDep, MeanF1, MeanF2, MeanF3: double;
minf1, maxf1, minf2, maxf2, minf3, maxf3, nofactors, totcells : integer;
cellcnts : DblDyneVec; // array of cell counts
cellcnts : IntDyneVec; // array of cell counts
cellvars : DblDyneVec; // arrray of cell sums of squares then variances
cellsums : DblDyneVec; // array of cell sums then means
equal_grp : boolean; // check for equal groups for post-hoc tests
counts : DblDyneMat; // matrix for 2-way containing cell sizes
counts : IntDyneMat; // matrix for 2-way containing cell sizes
sums : DblDyneMat; // matrix for 2-way containing cell sums
vars : DblDyneMat; // matrix for 2-way containing sums of squares
RowSums : DblDyneVec; // 2 way row sums
ColSums : DblDyneVec; // 2 way col sums
RowCount : DblDyneVec; // 2 way row count
ColCount : DblDyneVec; // 2 way col count
RowCount : IntDyneVec; // 2 way row count
ColCount : IntDyneVec; // 2 way col count
SlcSums : DblDyneVec; // 3 way slice sums
SlcCount : DblDyneVec; // 3 way slice counts
SlcCount : IntDyneVec; // 3 way slice counts
NoGrpsA, NoGrpsB, NoGrpsC : integer;
OrdMeansA, OrdMeansB, OrdMeansC : DblDyneVec; // reordered means for f1, f2, f3
allAlpha, PostHocAlpha : double; // alphas for tests
@ -534,7 +534,7 @@ end;
procedure TBlksAnovaFrm.GetLevels;
var
i: integer;
i, intValue: integer;
begin
Nf1cells := 0;
Nf2Cells := 0;
@ -545,7 +545,7 @@ begin
for i := 1 to NoCases do
begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col, i])));
if intvalue > maxf1 then maxf1 := intvalue;
if intvalue < minf1 then minf1 := intvalue;
end;
@ -558,7 +558,7 @@ begin
for i := 1 to NoCases do
begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F2Col,i])));
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F2Col, i])));
if intvalue > maxf2 then maxf2 := intvalue;
if intvalue < minf2 then minf2 := intvalue;
end;
@ -572,7 +572,7 @@ begin
for i := 1 to NoCases do
begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F3Col,i])));
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F3Col, i])));
if intvalue > maxf3 then maxf3 := intvalue;
if intvalue < minf3 then minf3 := intvalue;
end;
@ -586,7 +586,7 @@ end;
procedure TBlksAnovaFrm.Calc1Way;
var
i: integer;
i, intValue: integer;
X, X2: Double;
begin
CompError := false;
@ -597,11 +597,11 @@ begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
X := StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[DepVarCol,i]));
X := X*X;
intvalue := intvalue - minf1 + 1;
cellcnts[intvalue-1] := cellcnts[intvalue-1] + 1;
cellsums[intvalue-1] := cellsums[intvalue-1] + X;
cellvars[intvalue-1] := cellvars[intvalue-1] + X2;
X2 := X*X;
intvalue := intvalue - minf1;
cellcnts[intvalue] := cellcnts[intvalue] + 1;
cellsums[intvalue] := cellsums[intvalue] + X;
cellvars[intvalue] := cellvars[intvalue] + X2;
MeanDep := MeanDep + X;
SSDep := SSDep + X2;
N := N + 1;
@ -687,7 +687,7 @@ begin
end;
if cellcnts[i] > 0 then
AReport.Add('%4d %10.2f%10.2f%10.2f%4.0f', [
AReport.Add('%4d %10.2f%10.2f%10.2f%4d', [
i+1, cellsums[i] / cellcnts[i], cellvars[i], sqrt(cellvars[i]), cellcnts[i]
]);
end;
@ -761,152 +761,162 @@ var
Constant, RowsTotCnt, ColsTotCnt, SSCells : double;
X, X2: Double;
begin
CompError := false;
CompError := false;
// initialize matrix values
NoGrpsA := maxf1 - minf1 + 1;
NoGrpsB := maxf2 - minf2 + 1;
for i := 1 to NoGrpsA do
begin
RowSums[i-1] := 0.0;
RowCount[i-1] := 0.0;
for j := 1 to NoGrpsB do
begin
counts[i-1,j-1] := 0.0;
sums[i-1,j-1] := 0.0;
vars[i-1,j-1] := 0.0;
end;
end;
for i := 1 to NoGrpsB do
begin
ColCount[i-1] := 0.0;
ColSums[i-1] := 0.0;
end;
N := 0;
MeanDep := 0.0;
SSDep := 0.0;
SSCells := 0.0;
RowsTotCnt := 0.0;
ColsTotCnt := 0.0;
// get working totals
for i := 1 to NoCases do
begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
grpA := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
grpB := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F2Col,i])));
X := StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[DepVarCol,i]));
X := X*X;
grpA := grpA - minf1 + 1;
grpB := grpB - minf2 + 1;
counts[grpA-1,grpB-1] := counts[grpA-1,grpB-1] + 1;
sums[grpA-1,grpB-1] := sums[grpA-1,grpB-1] + X;
vars[grpA-1,grpB-1] := vars[grpA-1,grpB-1] + X2;
RowSums[GrpA-1] := RowSums[GrpA-1] + X;
ColSums[GrpB-1] := ColSums[GrpB-1] + X;
RowCount[GrpA-1] := RowCount[GrpA-1] + 1.0;
ColCount[GrpB-1] := ColCount[GrpB-1] + 1.0;
MeanDep := MeanDep + X;
SSDep := SSDep + X2;
N := N + 1;
end;
// initialize matrix values
NoGrpsA := maxf1 - minf1 + 1;
NoGrpsB := maxf2 - minf2 + 1;
for i := 0 to NoGrpsA-1 do
begin
RowSums[i] := 0.0;
RowCount[i] := 0;
for j := 0 to NoGrpsB-1 do
begin
counts[i, j] := 0;
sums[i, j] := 0.0;
vars[i, j] := 0.0;
end;
end;
// Calculate results
for i := 0 to NoGrpsA-1 do
begin
SSF1 := SSF1 + ((RowSums[i] * RowSums[i]) / RowCount[i]);
RowsTotCnt := RowsTotCnt + RowCount[i];
end;
for j := 0 to NoGrpsB-1 do
begin
SSF2 := SSF2 + ((ColSums[j] * ColSums[j]) / ColCount[j]);
ColsTotCnt := ColsTotCnt + ColCount[j];
end;
for i := 0 to NoGrpsA-1 do
begin
for j := 0 to NoGrpsB-1 do
if counts[i,j] > 0 then
SSCells := SSCells + ((sums[i,j] * sums[i,j]) / counts[i,j]);
end;
if N > 0 then Constant := (MeanDep * MeanDep) / N else Constant := 0.0;
SSF1 := SSF1 - Constant;
SSF2 := SSF2 - Constant;
SSF1F2 := SSCells - SSF1 - SSF2 - Constant;
SSErr := SSDep - SSCells;
SSDep := SSDep - Constant;
for i := 0 to NoGrpsB-1 do
begin
ColCount[i] := 0;
ColSums[i] := 0.0;
end;
if (SSF1F2 < 0) or (SSF1 < 0) or (SSF2 < 0) then
begin
MessageDlg('A negative SS found. Unbalanced design? Ending analysis.', mtError, [mbOK], 0);
CompError := true;
exit;
end;
N := 0;
MeanDep := 0.0;
SSDep := 0.0;
SSCells := 0.0;
RowsTotCnt := 0.0;
ColsTotCnt := 0.0;
DFTot := N - 1;
DFF1 := NoGrpsA - 1;
DFF2 := NoGrpsB - 1;
DFF1F2 := DFF1 * DFF2;
DFErr := DFTot - DFF1 - DFF2 - DFF1F2;
// DFCells := N - (NoGrpsA * NoGrpsB);
MSF1 := SSF1 / DFF1;
MSF2 := SSF2 / DFF2;
MSF1F2 := SSF1F2 / DFF1F2;
MSErr := SSErr / DFErr;
MSDep := SSDep / DFTot;
OmegaF1 := (SSF1 - DFF1 * MSErr) / (SSDep + MSErr);
OmegaF2 := (SSF2 - DFF2 * MSErr) / (SSDep + MSErr);
OmegaF1F2 := (SSF1F2 - DFF1F2 * MSErr) / (SSDep + MSErr);
Omega := OmegaF1 + OmegaF2 + OmegaF1F2;
MeanDep := MeanDep / N;
// f tests for fixed effects
if (Fact1Grp.ItemIndex = 0) and (Fact2Grp.ItemIndex = 0) then
begin
FF1 := abs(MSF1 / MSErr);
FF2 := abs(MSF2 / MSErr);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFErr);
ProbF2 := probf(FF2,DFF2,DFErr);
ProbF1F2 := probf(FF1F2,DFF1F2,DFErr);
end;
// f tests if both factors are random
if (Fact1Grp.ItemIndex = 1) and (Fact2Grp.ItemIndex = 1) then
begin
FF1 := abs(MSF1 / MSF1F2);
FF2 := abs(MSF2 / MSF1F2);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFF1F2);
ProbF2 := probf(FF2,DFF2,DFF1F2);
ProbF3 := probf(FF1F2,DFF1F2,DFErr);
end;
// f test if factor A is random
if (Fact1Grp.ItemIndex = 1) and (Fact2Grp.ItemIndex = 0) then
begin
FF1 := abs(MSF1 / MSErr);
FF2 := abs(MSF2 / MSF1F2);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFErr);
ProbF2 := probf(FF2,DFF2,DFF1F2);
ProbF3 := probf(FF1F2,DFF1F2,DFErr);
end;
// f test if factor b is random
if (Fact1Grp.ItemIndex = 0) and (Fact2Grp.ItemIndex = 1) then
begin
FF1 := abs(MSF1 / MSF1F2);
FF2 := abs(MSF2 / MSErr);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFF1F2);
ProbF2 := probf(FF2,DFF2,DFErr);
ProbF3 := probf(FF1F2,DFF1F2,DFErr);
end;
if (ProbF1 > 1.0) then ProbF1 := 1.0;
if (ProbF2 > 1.0) then ProbF2 := 1.0;
if (ProbF1F2 > 1.0) then ProbF1F2 := 1.0;
// get working totals
for i := 1 to NoCases do
begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
grpA := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
grpB := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F2Col,i])));
X := StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[DepVarCol,i]));
X := X*X;
grpA := grpA - minf1 + 1;
grpB := grpB - minf2 + 1;
counts[grpA-1,grpB-1] := counts[grpA-1,grpB-1] + 1;
sums[grpA-1,grpB-1] := sums[grpA-1,grpB-1] + X;
vars[grpA-1,grpB-1] := vars[grpA-1,grpB-1] + X2;
RowSums[GrpA-1] := RowSums[GrpA-1] + X;
ColSums[GrpB-1] := ColSums[GrpB-1] + X;
RowCount[GrpA-1] := RowCount[GrpA-1] + 1;
ColCount[GrpB-1] := ColCount[GrpB-1] + 1;
MeanDep := MeanDep + X;
SSDep := SSDep + X2;
N := N + 1;
end;
// Obtain omega squared (proportion of dependent variable explained)
if (OmegaF1 < 0.0) then OmegaF1 := 0.0;
if (OmegaF2 < 0.0) then OmegaF2 := 0.0;
if (OmegaF1F2 < 0.0) then OmegaF1F2 := 0.0;
//Omega = ( (SSF1 + SSF2 + SSF1F2) - (DFF1 + DFF2 + DFF1F2) * MSErr) / (SSDep + MSErr);
if (Omega < 0.0) then Omega := 0.0;
// Calculate results
for i := 0 to NoGrpsA-1 do
begin
SSF1 := SSF1 + ((RowSums[i] * RowSums[i]) / RowCount[i]);
RowsTotCnt := RowsTotCnt + RowCount[i];
end;
for j := 0 to NoGrpsB-1 do
begin
SSF2 := SSF2 + ((ColSums[j] * ColSums[j]) / ColCount[j]);
ColsTotCnt := ColsTotCnt + ColCount[j];
end;
for i := 0 to NoGrpsA-1 do
begin
for j := 0 to NoGrpsB-1 do
if counts[i,j] > 0 then
SSCells := SSCells + ((sums[i,j] * sums[i,j]) / counts[i,j]);
end;
if N > 0 then Constant := (MeanDep * MeanDep) / N else Constant := 0.0;
SSF1 := SSF1 - Constant;
SSF2 := SSF2 - Constant;
SSF1F2 := SSCells - SSF1 - SSF2 - Constant;
SSErr := SSDep - SSCells;
SSDep := SSDep - Constant;
if (SSF1F2 < 0) or (SSF1 < 0) or (SSF2 < 0) then
begin
ErrorMsg('A negative SS found. Unbalanced design? Ending analysis.');
CompError := true;
exit;
end;
DFTot := N - 1;
DFF1 := NoGrpsA - 1;
DFF2 := NoGrpsB - 1;
DFF1F2 := DFF1 * DFF2;
DFErr := DFTot - DFF1 - DFF2 - DFF1F2;
MSF1 := SSF1 / DFF1;
MSF2 := SSF2 / DFF2;
MSF1F2 := SSF1F2 / DFF1F2;
MSErr := SSErr / DFErr;
MSDep := SSDep / DFTot;
OmegaF1 := (SSF1 - DFF1 * MSErr) / (SSDep + MSErr);
OmegaF2 := (SSF2 - DFF2 * MSErr) / (SSDep + MSErr);
OmegaF1F2 := (SSF1F2 - DFF1F2 * MSErr) / (SSDep + MSErr);
Omega := OmegaF1 + OmegaF2 + OmegaF1F2;
MeanDep := MeanDep / N;
// F tests for fixed effects
if (Fact1Grp.ItemIndex = 0) and (Fact2Grp.ItemIndex = 0) then
begin
FF1 := abs(MSF1 / MSErr);
FF2 := abs(MSF2 / MSErr);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFErr);
ProbF2 := probf(FF2,DFF2,DFErr);
ProbF1F2 := probf(FF1F2,DFF1F2,DFErr);
end;
// F tests if both factors are random
if (Fact1Grp.ItemIndex = 1) and (Fact2Grp.ItemIndex = 1) then
begin
FF1 := abs(MSF1 / MSF1F2);
FF2 := abs(MSF2 / MSF1F2);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFF1F2);
ProbF2 := probf(FF2,DFF2,DFF1F2);
ProbF3 := probf(FF1F2,DFF1F2,DFErr);
end;
// F test if factor A is random
if (Fact1Grp.ItemIndex = 1) and (Fact2Grp.ItemIndex = 0) then
begin
FF1 := abs(MSF1 / MSErr);
FF2 := abs(MSF2 / MSF1F2);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFErr);
ProbF2 := probf(FF2,DFF2,DFF1F2);
ProbF3 := probf(FF1F2,DFF1F2,DFErr);
end;
// F test if factor b is random
if (Fact1Grp.ItemIndex = 0) and (Fact2Grp.ItemIndex = 1) then
begin
FF1 := abs(MSF1 / MSF1F2);
FF2 := abs(MSF2 / MSErr);
FF1F2 := abs(MSF1F2 / MSErr);
ProbF1 := probf(FF1,DFF1,DFF1F2);
ProbF2 := probf(FF2,DFF2,DFErr);
ProbF3 := probf(FF1F2,DFF1F2,DFErr);
end;
if (ProbF1 > 1.0) then ProbF1 := 1.0;
if (ProbF2 > 1.0) then ProbF2 := 1.0;
if (ProbF1F2 > 1.0) then ProbF1F2 := 1.0;
// Obtain omega squared (proportion of dependent variable explained)
if (OmegaF1 < 0.0) then OmegaF1 := 0.0;
if (OmegaF2 < 0.0) then OmegaF2 := 0.0;
if (OmegaF1F2 < 0.0) then OmegaF1F2 := 0.0;
if (Omega < 0.0) then Omega := 0.0;
end;
procedure TBlksAnovaFrm.TwoWayTable(AReport: TStrings);
@ -968,7 +978,7 @@ begin
AReport.Add('Descriptive Statistics');
AReport.Add('');
AReport.Add('GROUP Row Col. N MEAN VARIANCE STD.DEV.');
groupsize := round(counts[0,0]);
groupsize := counts[0, 0];
equal_grp := true;
MaxVar := 0.0;
MinVar := 1e20;
@ -1024,7 +1034,7 @@ begin
if ColCount[j] > 0 then V := ColSS - (ColSums[j] * ColSums[j] / ColCount[j]);
if ColCount[j] > 1 then V := V / (ColCount[j] - 1.0);
if V > 0.0 then S := sqrt(V);
AReport.Add('Col %3d %3.0f %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
AReport.Add('Col %3d %3d %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
end;
AReport.Add('TOTAL %3d %8.3f %8.3f %8.3f', [N, MeanDep, MSDep, sqrt(MSDep)]);
@ -1177,7 +1187,7 @@ begin
for i := 0 to NoGrpsA-1 do
begin
RowSums[i] := 0.0;
RowCount[i] := 0.0;
RowCount[i] := 0;
for j := 0 to NoGrpsB-1 do
begin
for k := 0 to NoGrpsC-1 do
@ -1191,13 +1201,13 @@ begin
for i := 0 to NoGrpsB-1 do
begin
ColCount[i] := 0.0;
ColCount[i] := 0;
ColSums[i] := 0.0;
end;
for i := 0 to NoGrpsC-1 do
begin
SlcCount[i] := 0.0;
SlcCount[i] := 0;
SlcSums[i] := 0.0;
end;
@ -1234,9 +1244,9 @@ begin
RowSums[GrpA-1] := RowSums[GrpA-1] + X;
ColSums[GrpB-1] := ColSums[GrpB-1] + X;
SlcSums[GrpC-1] := SlcSums[GrpC-1] + X;
RowCount[GrpA-1] := RowCount[GrpA-1] + 1.0;
ColCount[GrpB-1] := ColCount[GrpB-1] + 1.0;
SlcCount[GrpC-1] := SlcCount[GrpC-1] + 1.0;
RowCount[GrpA-1] := RowCount[GrpA-1] + 1;
ColCount[GrpB-1] := ColCount[GrpB-1] + 1;
SlcCount[GrpC-1] := SlcCount[GrpC-1] + 1;
MeanDep := MeanDep + X;
SSDep := SSDep + X2;
N := N + 1;
@ -1757,7 +1767,7 @@ begin
V := RowSS - (RowSums[i] * RowSums[i] / RowCount[i]);
V := V / (RowCount[i] - 1.0);
S := sqrt(V);
AReport.Add('Row %3d %3.0f %8.3f %8.3f %8.3f', [minf1+i, RowCount[i], XBar, V, S]);
AReport.Add('Row %3d %3d %8.3f %8.3f %8.3f', [minf1+i, RowCount[i], XBar, V, S]);
end;
//Display means, variances and standard deviations for columns
@ -1771,7 +1781,7 @@ begin
V := ColSS - (ColSums[j] * ColSums[j] / ColCount[j]);
V := V / (ColCount[j] - 1.0);
S := sqrt(V);
AReport.Add('Col %3d %3.0f %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
AReport.Add('Col %3d %3d %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
end;
//Display means, variances and standard deviations for slices
@ -1785,7 +1795,7 @@ begin
V := SlcSS - (SlcSums[k] * SlcSums[k] / SlcCount[k]);
V := V / (SlcCount[k] - 1.0);
S := sqrt(V);
AReport.Add('Slice %3d %3.0f %8.3f %8.3f %8.3f', [minf3+k, SlcCount[k], XBar, V, S]);
AReport.Add('Slice %3d %3d %8.3f %8.3f %8.3f', [minf3+k, SlcCount[k], XBar, V, S]);
end;
AReport.ADd('TOTAL %3d %8.3f %8.3f %8.3f', [N, MeanDep, MSDep, sqrt(MSDep)]);
@ -2447,7 +2457,7 @@ end;
//-------------------------------------------------------------------
procedure TBlksAnovaFrm.BrownForsytheOneWay(AReport: TStrings);
var
i: integer;
i, intValue: integer;
c1: array[1..50] of double;
cellmeans: array[1..50] of double;
sumc1: double;
@ -2464,12 +2474,12 @@ begin
for i := 1 to NoCases do
begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
// X := StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[DepVarCol,i]));
intvalue := intvalue - minf1 + 1;
cellcnts[intvalue-1] := 0.0;
cellsums[intvalue-1] := 0.0;
cellvars[intvalue-1] := 0.0;
intvalue := intvalue - minf1;
cellcnts[intvalue] := 0;
cellsums[intvalue] := 0.0;
cellvars[intvalue] := 0.0;
end;
MeanDep := 0.0;
@ -2484,10 +2494,10 @@ begin
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
X := StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[DepVarCol,i]));
X2 := X*X;
intvalue := intvalue - minf1 + 1;
cellcnts[intvalue-1] := cellcnts[intvalue-1] + 1;
cellsums[intvalue-1] := cellsums[intvalue-1] + X;
cellvars[intvalue-1] := cellvars[intvalue-1] + X2;
intvalue := intvalue - minf1;
cellcnts[intvalue] := cellcnts[intvalue] + 1;
cellsums[intvalue] := cellsums[intvalue] + X;
cellvars[intvalue] := cellvars[intvalue] + X2;
MeanDep := MeanDep + X;
SSDep := SSDep + X2;
N := N + 1;
@ -2581,7 +2591,7 @@ end;
procedure TBlksAnovaFrm.WelchOneWay(AReport: TStrings);
var
i: integer;
i, intValue: integer;
W, v, barx, numerator, denominator: double;
wj: array[1..50] of double;
c1: array[1..50] of double;
@ -2602,10 +2612,10 @@ begin
if not GoodRecord(i,NoSelected,ColNoSelected) then continue;
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
// X := StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[DepVarCol,i]));
intvalue := intvalue - minf1 + 1;
cellcnts[intvalue-1] := 0.0;
cellsums[intvalue-1] := 0.0;
cellvars[intvalue-1] := 0.0;
intvalue := intvalue - minf1;
cellcnts[intvalue] := 0;
cellsums[intvalue] := 0.0;
cellvars[intvalue] := 0.0;
end;
MeanDep := 0.0;
@ -2620,13 +2630,13 @@ begin
intvalue := round(StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[F1Col,i])));
X := StrToFloat(Trim(OS3MainFrm.DataGrid.Cells[DepVarCol,i]));
X := X*X;
intvalue := intvalue - minf1 + 1;
cellcnts[intvalue-1] := cellcnts[intvalue-1] + 1;
cellsums[intvalue-1] := cellsums[intvalue-1] + X;
cellvars[intvalue-1] := cellvars[intvalue-1] + X2;
intvalue := intvalue - minf1;
cellcnts[intvalue] := cellcnts[intvalue] + 1;
cellsums[intvalue] := cellsums[intvalue] + X;
cellvars[intvalue] := cellvars[intvalue] + X2;
MeanDep := MeanDep + X;
SSDep := SSDep + X2;
barxj[intvalue] := barxj[intvalue] + X;
barxj[intvalue+1] := barxj[intvalue+1] + X;
N := N + 1;
end;
@ -2770,8 +2780,8 @@ begin
denominator := sqrt((var1 / cellcnts[i-1]) + (var2 / cellcnts[j-1]));
numerator := mean1 - mean2;
t := numerator / denominator;
AReport.Add('Mean %d = %8.3f, Mean %d = %8.3f', [i, mean1, j, mean2]);
AReport.Add('Welch t = %8.3f' ,[t]);
AReport.Add('Mean %d: %8.3f, Mean %d: %8.3f', [i, mean1, j, mean2]);
AReport.Add('Welch t: %8.3f' ,[t]);
numerator := sqr((var1 /cellcnts[i-1]) + (var2 / cellcnts[j-1]));
term1 := sqr(var1) / (sqr(cellcnts[i-1]) * (cellcnts[i-1]-1.0));
@ -2779,7 +2789,7 @@ begin
denominator := term1 + term2;
numerator := sqr((var1 / cellcnts[i-1]) + (var2 / cellcnts[j-1]));
gnu := numerator / denominator;
AReport.Add('degrees of freedom = %8.3f', [gnu]);
AReport.Add('degrees of freedom: %8.3f', [gnu]);
v := round(gnu);
AReport.Add('Rounded degrees of freedom = %d', [v]);

44
applications/lazstats/source/forms/analysis/comparisons/onecaseanovaunit.pas
View File

@ -86,18 +86,18 @@ type
MeanDep, MeanF1, MeanF2, MeanF3: double;
SSNonAdd, SSBalance,MSNonAdd, MSBalance, GrandMean, DFBalance: double;
FNonAdd, ProbNonAdd: double;
cellcnts : DblDyneVec; // array of cell counts
cellcnts : IntDyneVec; // array of cell counts
cellvars : DblDyneVec; // arrray of cell sums of squares then variances
cellsums : DblDyneVec; // array of cell sums then means
counts : DblDyneMat; // matrix for 2-way containing cell sizes
counts : IntDyneMat; // matrix for 2-way containing cell sizes
sums : DblDyneMat; // matrix for 2-way containing cell sums
vars : DblDyneMat; // matrix for 2-way containing sums of squares
RowSums : DblDyneVec; // 2 way row sums
ColSums : DblDyneVec; // 2 way col sums
RowCount : DblDyneVec; // 2 way row count
ColCount : DblDyneVec; // 2 way col count
RowCount : IntDyneVec; // 2 way row count
ColCount : IntDyneVec; // 2 way col count
SlcSums : DblDyneVec; // 3 way slice sums
SlcCount : DblDyneVec; // 3 way slice counts
SlcCount : IntDyneVec; // 3 way slice counts
OrdMeansA, OrdMeansB, OrdMeansC : DblDyneVec; // reordered means for f1, f2, f3
OverAll, PostHocAlpha : double; // alphas for tests
wsum, wx2 : DblDyneCube; // : DblDyneCube
@ -544,10 +544,10 @@ begin
for i := 1 to NoGrpsA do
begin
RowSums[i-1] := 0.0;
RowCount[i-1] := 0.0;
RowCount[i-1] := 0;
for j := 1 to NoGrpsB do
begin
counts[i-1,j-1] := 0.0;
counts[i-1,j-1] := 0;
sums[i-1,j-1] := 0.0;
vars[i-1,j-1] := 0.0;
end;
@ -555,7 +555,7 @@ begin
for i := 1 to NoGrpsB do
begin
ColCount[i-1] := 0.0;
ColCount[i-1] := 0;
ColSums[i-1] := 0.0;
end;
@ -584,8 +584,8 @@ begin
vars[grpA-1,grpB-1] := vars[grpA-1,grpB-1] + X * X;
RowSums[grpA-1] := RowSums[grpA-1] + X;
ColSums[grpB-1] := ColSums[grpB-1] + X;
RowCount[grpA-1] := RowCount[grpA-1] + 1.0;
ColCount[grpB-1] := ColCount[grpB-1] + 1.0;
RowCount[grpA-1] := RowCount[grpA-1] + 1;
ColCount[grpB-1] := ColCount[grpB-1] + 1;
MeanDep := MeanDep + X;
SSDep := SSDep + X * X;
N := N + 1;
@ -743,7 +743,7 @@ begin
V := RowSS - sqr(RowSums[i]) / RowCount[i];
V := V / (RowCount[i] - 1.0);
S := sqrt(V);
AReport.Add('Row %3d %3.0f %8.3f %8.3f %8.3f', [minf1+i, RowCount[i], XBar, V, S]);
AReport.Add('Row %3d %3d %8.3f %8.3f %8.3f', [minf1+i, RowCount[i], XBar, V, S]);
end;
//Display means, variances and standard deviations for columns
@ -756,7 +756,7 @@ begin
if (ColCount[j] > 0) then V := ColSS - sqr(ColSums[j]) / ColCount[j];
if (ColCount[j] > 1) then V := V / (ColCount[j] - 1.0);
if (V > 0.0) then S := sqrt(V);
AReport.Add('Col %3d %3.0f %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
AReport.Add('Col %3d %3d %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
end;
AReport.Add('TOTAL %3d %8.3f %8.3f %8.3f', [N, MeanDep, MSDep, sqrt(MSDep)]);
@ -899,7 +899,7 @@ begin
for i := 0 to NoGrpsA - 1 do
begin
RowSums[i] := 0.0;
RowCount[i] := 0.0;
RowCount[i] := 0;
for j := 0 to NoGrpsB - 1 do
begin
for k := 0 to NoGrpsC - 1 do
@ -913,14 +913,14 @@ begin
for i := 0 to NoGrpsB - 1 do
begin
ColCount[i] := 0.0;
ColCount[i] := 0;
ColSums[i] := 0.0;
end;
for i := 0 to NoGrpsC - 1 do
begin
SlcCount[i] := 0.0;
SlcSums[i] := 0.0;
SlcCount[i] := 0;
SlcSums[i] := 0;
end;
N := 0;
@ -956,9 +956,9 @@ begin
RowSums[grpA-1] := RowSums[grpA-1] + X;
ColSums[grpB-1] := ColSums[grpB-1] + X;
SlcSums[grpC-1] := SlcSums[grpC-1] + X;
RowCount[grpA-1] := RowCount[grpA-1] + 1.0;
ColCount[grpB-1] := ColCount[grpB-1] + 1.0;
SlcCount[grpC-1] := SlcCount[grpC-1] + 1.0;
RowCount[grpA-1] := RowCount[grpA-1] + 1;
ColCount[grpB-1] := ColCount[grpB-1] + 1;
SlcCount[grpC-1] := SlcCount[grpC-1] + 1;
MeanDep := MeanDep + X;
SSDep := SSDep + X * X;
N := N + 1;
@ -1222,7 +1222,7 @@ begin
V := RowSS - (RowSums[i] * RowSums[i] / RowCount[i]);
V := V / (RowCount[i] - 1.0);
S := sqrt(V);
AReport.Add('Row %3d %3.0f %8.3f %8.3f %8.3f', [minf1+i, RowCount[i], XBar, V, S]);
AReport.Add('Row %3d %3d %8.3f %8.3f %8.3f', [minf1+i, RowCount[i], XBar, V, S]);
end;
//Display means, variances and standard deviations for columns
@ -1237,7 +1237,7 @@ begin
V := ColSS - (ColSums[j] * ColSums[j] / ColCount[j]);
V := V / (ColCount[j] - 1.0);
S := sqrt(V);
AReport.Add('Col %3d %3.0f %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
AReport.Add('Col %3d %3d %8.3f %8.3f %8.3f', [minf2+j, ColCount[j], XBar, V, S]);
end;
//Display means, variances and standard deviations for slices
@ -1252,7 +1252,7 @@ begin
V := SlcSS - (SlcSums[k] * SlcSums[k] / SlcCount[k]);
V := V / (SlcCount[k] - 1.0);
S := sqrt(V);
AReport.Add('Slice %3d %3.0f %8.3f %8.3f %8.3f', [minf3+k, SlcCount[k], XBar, V, S]);
AReport.Add('Slice %3d %3d %8.3f %8.3f %8.3f', [minf3+k, SlcCount[k], XBar, V, S]);
end;
AReport.Add('TOTAL %3d %8.3f %8.3f %8.3f', [N, MeanDep, MSDep, sqrt(MSDep)]);

43
applications/lazstats/source/units/anovatestsunit.pas
View File

@ -18,7 +18,7 @@ procedure Tukey(
error_df : double; { deg. freedom for residual }
value : double; { size of smallest group }
group_total : DblDyneVec; { sum of scores in a group }
group_count : DblDyneVec; { no. of cases in a group }
group_count : IntDyneVec; { no. of cases in a group }
min_grp : integer; { minimum group code }
max_grp : integer; { maximum group code }
Alpha : Double; { alpha value }
@ -27,7 +27,7 @@ procedure Tukey(
procedure ScheffeTest(
error_ms : double; { mean squared residual }
group_total : DblDyneVec; { sum of scores in a group }
group_count : DblDyneVec; { count of cases in a group }
group_count : IntDyneVec; { count of cases in a group }
min_grp : integer; { code of first group }
max_grp : integer; { code of last group }
total_n : double; { total number of cases }
@ -39,7 +39,7 @@ procedure Newman_Keuls(
error_df : double; { deg. freedom for error }
value : double; { number in smallest group }
group_total : DblDyneVec; { sum of scores in a group }
group_count : DblDyneVec; { count of cases in a group }
group_count : IntDyneVec; { count of cases in a group }
min_grp : integer; { lowest group code }
max_grp : integer; { largest group code }
Alpha : double; { alpha value for testing }
@ -50,7 +50,7 @@ procedure Tukey_Kramer(
error_df : double; { deg. freedom for error }
value : double; { number in smallest group }
group_total : DblDyneVec; { sum of scores in group }
group_count : DblDyneVec; { number of caes in group }
group_count : IntDyneVec; { number of caes in group }
min_grp : integer; { code of lowest group }
max_grp : integer; { code of highst group }
Alpha : double; { Alpha value for testing }
@ -60,7 +60,7 @@ procedure Contrasts(
error_ms : double; { residual ms }
error_df : double; { residual df }
group_total : DblDyneVec; { group sums }
group_count : DblDyneVec; { group cases }
group_count : IntDyneVec; { group cases }
min_grp : integer; { lowest code }
max_grp : integer; { highest code }
overall_probf : double; { prob of overall test }
@ -68,7 +68,7 @@ procedure Contrasts(
procedure Bonferroni(
group_total : DblDyneVec; { sum of scores in group }
group_count : DblDyneVec; { number of caes in group }
group_count : IntDyneVec; { number of caes in group }
group_var : DblDyneVec; { group variances }
min_grp : integer; { code of lowest group }
max_grp : integer; { code of highst group }
@ -79,7 +79,7 @@ procedure TukeyBTest(
ErrorMS : double; { within groups error }
ErrorDF : double; { degrees of freedom within }
group_total : DblDyneVec; { vector of group sums }
group_count : DblDyneVec; { vector of group n's }
group_count : IntDyneVec; { vector of group n's }
min_grp : integer; { smallest group code }
max_grp : integer; { largest group code }
groupsize : double; { size of groups (all equal) }
@ -102,7 +102,7 @@ procedure Tukey(error_ms : double; { mean squared for residual }
error_df : double; { deg. freedom for residual }
value : double; { size of smallest group }
group_total : DblDyneVec; { sum of scores in a group }
group_count : DblDyneVec; { no. of cases in a group }
group_count : IntDyneVec; { no. of cases in a group }
min_grp : integer; { minimum group code }
max_grp : integer; { maximum group code }
Alpha : double; { alpha value }
@ -155,7 +155,7 @@ end;
procedure ScheffeTest(error_ms : double; { mean squared residual }
group_total : DblDyneVec; { sum of scores in a group }
group_count : DblDyneVec; { count of cases in a group }
group_count : IntDyneVec; { count of cases in a group }
min_grp : integer; { code of first group }
max_grp : integer; { code of last group }
total_n : double; { total number of cases }
@ -209,14 +209,15 @@ procedure Newman_Keuls(error_ms : double; { residual mean squared }
error_df : double; { deg. freedom for error }
value : double; { number in smallest group }
group_total : DblDyneVec; { sum of scores in a group }
group_count : DblDyneVec; { count of cases in a group }
group_count : IntDyneVec; { count of cases in a group }
min_grp : integer; { lowest group code }
max_grp : integer; { largest group code }
alpha : double; { alpha value for testing }
AReport : TStrings);
var
i, j : integer;
temp1, temp2 : double;
i, j: integer;
temp1: double;
temp2: Integer;
groupno : IntDyneVec;
contrast, mean1, mean2 : double;
q_stat : double;
@ -232,8 +233,7 @@ begin
begin
for j := i + 1 to max_grp do
begin
if group_total[i-1] / group_count[i-1] >
group_total[j-1] / group_count[j-1] then
if group_total[i-1] / group_count[i-1] > group_total[j-1] / group_count[j-1] then
begin
temp1 := group_total[i-1];
temp2 := group_count[i-1];
@ -293,7 +293,7 @@ procedure Tukey_Kramer(error_ms : double; { residual mean squared }
error_df : double; { deg. freedom for error }
value : double; { number in smallest group }
group_total : DblDyneVec; { sum of scores in group }
group_count : DblDyneVec; { number of caes in group }
group_count : IntDyneVec; { number of caes in group }
min_grp : integer; { code of lowest group }
max_grp : integer; { code of highst group }
Alpha : Double; { Alpha value for testing }
@ -341,10 +341,10 @@ end;
{ ------------------------------------------------------------------------ }
procedure Contrasts(error_ms : double; { residual ms }
error_df : double; { residual df }
procedure Contrasts(error_ms : double; { residual ms }
error_df : double; { residual df }
group_total : DblDyneVec; { group sums }
group_count : DblDyneVec; { group cases }
group_count : IntDyneVec; { group cases }
min_grp : integer; { lowest code }
max_grp : integer; { highest code }
overall_probf : double; { prob of overall test }
@ -449,7 +449,7 @@ end; { of procedure CONTRASTS }
procedure Bonferroni( group_total : DblDyneVec; { sum of scores in group }
group_count : DblDyneVec; { number of cases in group }
group_count : IntDyneVec; { number of cases in group }
group_var : DblDyneVec; { group variances }
min_grp : integer; { code of lowest group }
max_grp : integer; { code of highst group }
@ -502,7 +502,7 @@ end;
procedure TukeyBTest(ErrorMS : double; { within groups error }
ErrorDF : double; { degrees of freedom within }
group_total : DblDyneVec; { vector of group sums }
group_count : DblDyneVec; { vector of group n's }
group_count : IntDyneVec; { vector of group n's }
min_grp : integer; { smallest group code }
max_grp : integer; { largest group code }
groupsize : double; { size of groups (all equal) }
@ -514,7 +514,8 @@ var
qstat: double;
tstat: double;
groupno: IntDyneVec;
temp1, temp2: double;
temp1: Double;
temp2: Integer;
tempno: integer;
NoGrps: integer;
contrast: double;