Author Topic: Kinetics - the RI factor and Chi2, any recommendations?  (Read 802 times)

GE Asa

  • Newbie
  • *
  • Posts: 19
Kinetics - the RI factor and Chi2, any recommendations?
« on: September 23, 2015, 07:15:21 AM »
When is my fit good?
 
This is how I, no big fan of equations, usually think around this. There is no general recommendation for Chi2 range. Chi2 is in RU2 and should be judged in relation to the response levels. If all data are within 20 RU and Chi2 is 16 RU2 then Chi2 is significant (square root of 16 is 4 compared to 20 RU = 20 % average deviation in the fit). However a Chi2 of 16 RU2 may be ok for 200 RU response level data (4 compared to 200 RU = 2 % average deviation).

If Chi2 is high look also at the residuals and judge if residuals are evenly distributed or if there are large local deviations perhaps caused by disturbances that could be removed.

I see 4 different scenarios for handling of RI factor:

1)   If the association phase and the dissociation phase are nicely connected at the end of injection when looking at the experimental data (the dissociation is starting immediately at the top of the sensorgram at end of injection without any visible “jump”), then no bulk shifts (RI shifts) exist in the experimental data and the RI factor should be set to zero before fitting (in Parameters, RI set to Constant, 0).

2)   If responses consists of square injections (no visible dissociation) then RI should always be set to 0.

3)   This is the tricky one...If it is hard to judge by looking at the sensorgram whether there is a bulk shift in the experimental data or not then RI default setting is used, “Fit local”. After fitting Chi2 may be low and if also RI values are reasonably low in relation to the response levels then all is usually well and the sun is shining and so are you! But if Chi2 is low and instead RI values are high and significant compared to the response levels (as read in the Report table or visualized in the Tools/Components plot), then the fit is still not good. Look at raw data, are bulk levels very high so that large values are subtracted in the reference subtraction? If so try to reduce bulks, normally 100 -200 RU bulks are subtracted without problems. Also check blank subtractions.

4) If there are visible jumps at beginning and end of injections then RI default setting is also used.

Please add any thoughts you may have on this wonderful world of trying to understand your data  :)

/Asa

« Last Edit: September 23, 2015, 07:23:46 AM by GE Asa »