dfittool - normal distribution plot

I have plottet my data with normal distribution using "dfittool" on matlab and it looks like below:

The thing is, i really dont understand what these strange peaks means ?? Is it the probability which tells, so many % of the data is located at this point ?

Thanks for your time.

[dpb]

Looks like it means you've got a sampled data set that includes about the expected numbers of observations per bin but for two binnings for the two sets.

Would need to know how you got the data into the tool and what those data are to be completely certain, of course...

[Morteza]

The data are 3 months voltage quality observation of a household. This is a 10 minute data, that means 6 masurements each hour.

[Youssef Khmou]

This result indicates that the data is normal random variable with mean of 229, but maybe the data contains some irregularities, if you try with x=229+randn(700,1) there will be no strange peaks....

[José-Luis]

No, the plot does not show that. It shows how the data compare to a normal distribution. You'd need a Kolmogorov-Smirnov or something similar to test the hypothesis that the data comes from a normal distribution.

Also, @Youssef, what would be the ultimate purpose of generating random data from a normal distribution and then fitting to a normal distribution, as you seem to be suggesting? Granted, no peaks, but then you are not measuring anything. Note also, that the variance is probably not what you set it to be.

[Youssef Khmou]

correct, but the majority of simulations are based on N(m,s)...

[José-Luis]

I'm not sure that's true just from that plot.

[dpb]

@Morteza...I think there's something else going on in the data set than just a single variable at 10-min observations from what this plot shows. It looks like two different distributions that have been somehow lumped.

The one w/ the higher values is only about 15-16 bins and isn't very normal at all--it's skewed strongly to the left while the other data are still left-skewed but not quite so strongly it appears, perhaps, and at a much finer binning.

The joint of the two fit by the estimated mean/variance is, as can be observed, of a mean location that reflects the left-skewness by the larger area above the bulk of the histogram on the left and the higher peak value than the max if the solid-looking histogram owing to the existence of the other values of fewer bins but much more magnitude when normalized.

Again, need to see a representative sample of the actual data to be able to tell what's actually going on behind the scenes.