7 Replies Latest reply: Nov 16, 2013 10:18 AM by Stefan Wühl RSS

    Distribution normal

    Iñigo Elviro Bodoy

      Hello everybody,

       

      I have a question about the normal distribution.

       

      I can not get a chart completly simetric what it is the usual in this kind of chart:

       

      1.png

      The formuls are:

       

      Dimensions: class([Net Weight Roll Control (gr)],0.5)

       

      Expressions:


      normdist( (floor(Min([Net Weight Roll Control (gr)]),0.5)+Ceil(Min([Net Weight Roll Control (gr)]),0.5))/2,

      avg({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),

      stdev({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),0)

       

       

      Count({<[Net Weight Roll Control (gr)]={">0"}>}[Net Weight Roll Control (gr)])

       

      What I try with: floor(Min([Net Weight Roll Control (gr)]),0.5)+Ceil(Min([Net Weight Roll Control (gr)]),0.5))/2 is to fix the medium of each interval to draw the distribution. In theory, it must be simetric, but follows the form got in the frecuency formul.

       

      I attached the excell file.

       

      Please, don´t send me the answer in qlik examples, because i have edition personal.

       

      Thanks in advance

        • Re: Distribution normal
          Stefan Wühl

          That's because your x-axis is not continuous. Have a close look to the intervals shown at the axis, you'll notice that some intermediate intervals are missing (e.g.118 - 118.5).

           

          If you use a continuous axis on axis tab, the curve should look much more as you expect it.

          • Re: Distribution normal
            Clever Anjos

            Using swuehl recomendation

              • Re: Distribution normal
                Iñigo Elviro Bodoy

                Hi Clever:

                 

                But I am usign Persoanl Edition, therefore I can not open the file.

                 

                Could you explain what is your solution with formuls?

                 

                Thanks in advance

                  • Re: Re: Distribution normal
                    Clever Anjos

                    I´ve used an expression (and marked as Line/Smooth)

                    normdist( (floor(Min([Net Weight Roll Control (gr)]),0.5)+Ceil(Min([Net Weight Roll Control (gr)]),0.5))/2,

                    avg({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),

                    stdev({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),0)

                     

                    I marked "continuous" into Axes Tab

                    Capturar.PNG.png

                     

                    419564.png

                      • Re: Distribution normal
                        Clever Anjos

                        normdist( Count({<[Net Weight Roll Control (gr)]={">0"}>}[Net Weight Roll Control (gr)]),

                        avg({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),

                        stdev({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),0)

                         

                        gives almost the same curve

                          • Re: Distribution normal
                            Iñigo Elviro Bodoy

                            Hi Clever:

                             

                            Thanks for your answer but the chart is not completely simetric. I have attached your chart and my chart where is more evident:

                            1.jpeg.jpg2.jpeg.jpg

                            The problem is the lack of values in the midle of the intervals.

                             

                            For solving this problem is necessary to solve this lack.

                             

                            Any answer more?

                             

                            Thanks in advance,

                              • Re: Re: Distribution normal
                                Stefan Wühl

                                I think you won't necessarily get a perfect symmetric curve because you are using discrete sampling points.

                                 

                                You can get a smoother, more symmetric curve if you create additional dimension values so your line curve gets calculated for more sampling points:

                                 

                                // Script to read your data and create a complete range of dimension values (adapt 115 start value and 20 sampling

                                // points to your range

                                 

                                LOAD class(115+recno()/2,0.5) as Class

                                AutoGenerate 20;

                                 

                                LOAD [Net Weight Roll Control (gr)],

                                  class([Net Weight Roll Control (gr)],0.5) as Class

                                FROM

                                [Net Values Weight.xlsx]

                                (ooxml, embedded labels, table is Sheet1);

                                 

                                Then use a dimension Class and as expression for the line

                                 

                                =normdist( Class,

                                avg({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),

                                stdev({1} total {<[Net Weight Roll Control (gr)]={">0"}>} [Net Weight Roll Control (gr)]),0)