Combination of two algorithms
viralvector
I try to compare several groups to a common control. I have been trying to add a customized algorithm mean squared error (see the image ) into Dunnett Test function but failed.
Can you give me an idea about combining these two algorithm togather?
Here is the mean squared error code:
Function WSFI(ctrlwave, treatwave, xl, xr) wave ctrlwave, treatwave //control waves variable xl, xr // (left and right bracket) duplicate/R=(xl, xr)/FREE bw, bdw duplicate/R=(xl, xr)/FREE tw, tdw MatrixOP/FREE MSE=Sum(magsqr(bdw - tdw))/numPoints(tdw) return MSE[0] End
Here is the DunnettTest function which was written by my boss
Function pksDunnettFn(ctrlWave) String ctrlWave WAVE M_DunnettTestResults WAVE /T T_DunnettDescriptors //Prompt ctrlWave,"Select control wave:",popup,WaveList("*",";","")+";_none_" String wName, wList, waveLabel, nStars Variable j, resultNrows, pValue, qValue, nWaves, totalSamples // print ctrlWave // compile wave name list from top table: ctrl wave first element wList = ctrlWave + ";" j = 0 totalSamples = numpnts($ctrlWave) // cumulate total # of samples do wName = WaveName("", j, 1) wName = ReplaceString(".d", wName, "") // strip '.d' from wave names sometimes if ( (cmpstr(wName ,"") != 0) && (cmpstr(wName ,ctrlWave) != 0) ) // append wave name if not null or ctrlWave name wList = wList + wName + ";" totalSamples += numpnts($wName) // cumulate total # of samples endif j += 1 while (cmpstr(wName ,"") != 0 ) nWaves = j-1 //print nWaves print wList // do the Dunnett's // StatsDunnettTest /ALPH=0.05 /T=1 /WSTR=wList /TAIL=4 /SWN StatsDunnettTest /ALPH=0.05 /WSTR=wList /TAIL=4 /SWN resultNrows = DimSize(M_DunnettTestResults, 0 ) // # rows in result Make /T /O /N=(resultNrows+1, 5) pksDunnettResultTable pksDunnettResultTable[0][0] = "CONTROL" pksDunnettResultTable[0][1] = "TEST GRP" pksDunnettResultTable[0][3] = "CONTROL MEAN" pksDunnettResultTable[0][4] = "TEST GRP MEAN" pksDunnettResultTable[0][2] = "P VALUE *0.05 **0.01" for (j=0; j<resultNrows; j+=1) // waveLabel = GetDimLabel(M_DunnettTestResults, 0, j ) // next test wave // wName = StringFromList(j+1, wList) waveLabel = T_DunnettDescriptors[j] wName = StringFromList(1, ReplaceString(" vs ", waveLabel,";")) pValue = M_DunnettTestResults[j][5] if (pValue >= 0.05) nStars = "" elseif (pValue >= 0.01) nStars = "*" else nStars = "**" endif if ( M_DunnettTestResults[j][4] == 0) // print ctrlWave + " vs " + wName + ": mean " + ctrlWave + " (control) = " + num2str(mean($ctrlWave)) + ", mean " + wName + " = " + num2str(mean($wName)) + ". Means ARE different (P = " + num2str(pValue) + ")" + nStars else // print ctrlWave + " vs " + wName + ": mean " + ctrlWave + " (control) = " + num2str(mean($ctrlWave)) + ", mean " + wName + " = " + num2str(mean($wName)) + ". Means are NOT different (P = " + num2str(pValue) + ")" endif pksDunnettResultTable[j+1][0] = ctrlWave pksDunnettResultTable[j+1][3] = num2str(mean($ctrlWave)) pksDunnettResultTable[j+1][1] = wName pksDunnettResultTable[j+1][4] = num2str(mean($wName)) pksDunnettResultTable[j+1][2] = num2str(pValue) + " " + nStars endfor print "* = significant at 5% level; ** = significant at 1% level" Edit /K=1 /W=(400, 70, 400+600, 300)pksDunnettResultTable as "Dunnett Result Table" ModifyTable /Z width[1]=140, width[2]=140, width[3]=140, showParts=16+32+64 end function
May 6, 2014 at 02:51 pm - Permalink
I am sorry that I did not clarify the goal. Yet you generously gave the best guesses for the goal.
I tried to calculate the mean square error within a range of waves used for Dunnett's test.
I have two groups for Dunnett's test
group 1- control
group 2- different treatments. eg. 5 different waves from 5 different treatments. group2a-2e
I can easily perform Dunnett's test, but I want to add mean square error into it.
At the end, I will have such data of comparison:
group1- group2a = Dunnett's score + mean square error
group1- group2b =Dunnett's score + mean square error
group1- group2c =Dunnett's score + mean square error
group1- group2d =Dunnett's score + mean square error
group1- group2e =Dunnett's score + mean square error
The mean square error which was written here ( the help from IGOR team previously)
xl and xr are the bracket to set the range of wave.
supposedly, ctrlwave is the ctrlwave shares with Dunnett's test; treatwave is the group of 2a-2e.
I tried your suggestion, but failed to compile it with the message saying:
Got "WSFI" instead of string variable or string function name.
Error in Stats:pksDunnettFn
pksDunnettResultTable[j+1][5] = WSFI (cw,ww,xl,xr)
Any clues?
Thank you very much!!
May 9, 2014 at 10:31 am - Permalink
change
pksDunnettResultTable[j+1][5] = WSFI(cw, ww, xl, xr)to
pksDunnettResultTable[j+1][5] = num2str( WSFI(cw, ww, xl, xr) )The former line attempts to assign a numeric value to a text wave. The latter converts the numeric value to a string before doing the assignment.
May 9, 2014 at 12:32 pm - Permalink
Thank you for the input. The error is gone.
I have received another error after running the function.
I got :
Function Execution Error
While executing Make, the following error occurred:
The number of points in wave must be between 0 and 2147 million.
and the table show up with empty value.
....Sot sure where to solve it.
Thank you in advance.
May 9, 2014 at 04:00 pm - Permalink
Read about the debugger:
DisplayHelpTopic "The Debugger"
John Weeks
WaveMetrics, Inc.
support@wavemetrics.com
May 9, 2014 at 04:19 pm - Permalink
The only statement that I can see involving the Make operation is this one
Make /T /O /N=(resultNrows+1, 6) pksDunnettResultTable // @@change 1Either use the debugger as John suggests or a print statement to see what value resultNrows is taking on.
The debugger is a very useful tool to have. You should follow John's advice.
May 9, 2014 at 07:36 pm - Permalink