What is the correct way to normalized spectral data?
I have a series of spectra. I have summed the data points within certain bins to create a histogram which represents the peaks I am interested in. I wish to compare these bins meaningfully by normalizing them.
For each bin, I have tried dividing by two factors:
1) Divide by the sum total of all bins (for one spectrum)
2) Divide by the mean of all bins (for one spectrum)
When graphed, the resulting normalized bins using both methods appear identical. I’m not sure if it’s the correct way, however.
What is the scientifically accepted way to normalize data?
Observing members:
0
Composing members:
0
5 Answers
This is an extremely specific question, and I think you would be better off consulting a textbook, talking to a college professor or other learned personage, or maybe just Googling this and interpreting the results. (You’re a college kid, right?)
Without knowing what your experiment is, it is hard to give you specific advice. First, why do you want to normalize? Second, do you have a negative control trace that you can subtract (as baseline) or normalize to? Another way it set either the minimum or maximum value to 1, and divide all the remaining values by that to normalize. Perhaps with more specific information, I can help more.
Normalizing data means to express the data in terms of the mean and standard deviation of the data. In this way, both sets of data are put on the same scale. This scale has zero in the center, -2 or -3 standard deviations on the left, and +2 or+3 standard deviations on the right.
You normalize by first finding the mean and standard deviation of the whole data set. Then you find the location of each data point in relation to its standard deviation.
For example, you have 50 data points, they average 100 with a standard deviation of 10. then 112 would be 12 above the mean or 1.2 standard deviations above the mean so that point would be at 1.2. 95 would be 5 points below the mean, so it would be -½ std dev, so it would be plotted at -.5. etc.
@PerryDolia, ok so I take the mean and the standard deviation (stdev).
If I divide each data point by the stdev, that makes the data set univariate, is that right?
And then if I subtract the mean from each data point, that makes the data set mean centered?
And together, these two operations make the data set normalized? In other words, I should first subtract the mean from each data point and then divide it by the stdev, right?
@PerryDolia, I just tried representing the data the way I described above as a column graph. Some bins are above the mean-center and some are below. Does this mean that those bins are anti-correlated?
Answer this question 
This question is in the General Section. Responses must be helpful and on-topic.
Composing members:
0
