Baseline - Science topic

Hello,

I am also facing sensitivity issue in PRM method. Would you please share how you fixed it?

In my personal opinion, it is possible to compare 5-min baseline HRV against 10-min task HRV. Even shorter durations of 1-min has been suggested to correlate with 5-min segments (PMID: 34211411). However, whether it is technically correct to compare 5-min vs 10-min segments will be dependent on whether expected HRV changes with your task is time dependent (do you expect difference between first 5 and last 5-min of the task?). Without knowing the exact task at hand, it may be difficult to conclude.

If unsure, I'd suggest to segregate post-task HRV in 5-min blocks and analyze to see if any differences exist (such as has been performed in PMID: 33300852). In any case, peer review may still request this information to justify the differences in analysis durations pre- and post-task.

Do you have any citations??

William Letter Thank you so much for your resources

Administer same test at the beginning (pre-test) and do same at the end (post-test) and find the difference between pre-test and post-test.

Thanks a lot for your thoughts, Apostolos. I appreciate it. I guess I'll have to do some post-processing then.

Just a few thoughts. 1. The baseline should be calibrated within the temperature range extended beyond the range of your runs. 2. While 600C is below melting point of your alloys, it is possible that at such high temperature aluminum pans may deform or react with the alloy. 3. May be, you should try performing your baseline calibrations and runs using platinum or ceramic pans.

Could be a bubble stuck in the conductivity meter.

Kindly use automatic baseline opition, if you have then you can set it manually and that may be helpful.

The software allows you to opt for analysis with or without ROX. however, it is highly recommended that you use high-ROX when supported by your RT-Thermocycling machine.

The OCP means that there is no change in the potential by injecting feed into the MFC, so the system is stable. This could be evaluated when two consecutive peaks (potential) appear.

You can also take a look at this paper for more illustration (Fig. 2):

First, blank the spectrophotometer. You do this with a solution that has ALL of the components that you would expect to find in your sample EXCEPT the molecule of interest. So, if you want to measure antibody that is in PBS/EDTA your blank has to be PBS/EDTA.

However you should also consider how the sample has been prepared. If you desalted the antibody into the PBS/EDTA buffer, the blank is the column equilibration buffer. If you dialysed the Ab into buffer, the strictly correct blank to use is the spent dialysis buffer (not fresh buffer). If you diluted 10mg/ml Ab in 50 mM Tris/0.1 % azide into PBS/EDTA to give a 1mg/ml Ab solution, the strictly correct blank to use then is a 1/10 dilution of the tris/azide buffer in the PBS/EDTA buffer, so that the ONLY difference between the blank and sample is the presence of the molecule that you wish to measure.

Finally, you will also need to use a cuvette that can transmit UV wavelengths e.g., quartz.

Speaking in general terms, a baseline subtraction is removing whatever contribution your data has that is completely irrelevant to your data analysis. I can only say something about Raman, but the best way to subtract a baseline is to measure your sample and your substrate using the same laser power, number of acquistions and time and subtract, from the sample signal, the signal from your substrate. If your sample has some contribution coming from some sort of luminescence, one way to remove the luminescent background would be to measure your sample in a parallel polarization configuration and follow that with a cross polarization configuration. Generally, the PL is unpolarized, and your measurement in cross polarization should remove the totally symmetric Raman modes, leaving only the other types of modes and the PL, which you can use to remove the PL from the parallel configuration. Again, if you know enough about Raman, and even PL, you'll know that these are suggestions rather than rules, because there is not a one perfect rule for all cases. As with most people though, you'll eventually have enough experience to do background subtraction easily.

Hope that helps.

I have done this test many times. As far as I know, it doesn't matter much if you do it under a homogeneous light condition and likewise test all the experimental groups at about the same time. For example, testing the control group at 8 am and the experimental group at 3 pm can make a difference depending on the circadian rhythm. After all, since you will be comparing the experimental groups with the control group, it will be sufficient to do it at the same time.

Yes, it should change when the amount of less volatile liquid, water, is added to methanol. What you have observed is normal. Water is far less volatile than Methanol and take different nebulizer, heat and gas flow settings. Each mobile phase must be evaluated to determine the best ELSD settings for use in your application. The signal to noise (S/N) level of a standard peak must be evaulated with each setting change to find the optimized conditions for use (To evaluate "noise", never rely on the "baseline" signal only! Often, as the baseline noise changes, so does the signal level, so monitoring the baseline is unreliable. S/N only of signal). Failure to optimize the ELSD to each mobile phase used may result in instability of signal and internal contamination. Once the settings are optimized, autozero the signal and monitor to determine ripple.

Please find a professional chromatographer with practical experience using HPLC and esp ELSD detectors in your area for to assist you. To use an ELSD, you must first receive professional training in how to use this complex HPLC detector (ELSD). Few have experience using them and you can not learn how to use one in a day. This detector is not a simple detector like an RID or UV/VIS is. An ELSD module is one of the more complex detectors you can use (similar to a MS or MS/MS system requiring many years of training).

We only recommend ELSD/CAD for advanced level users (> 10 years professional exp) as they are very difficult to use and obtain reliable data. *Be cautious of the sales people/vendors who provide misinformation about them to make a sale.

Just a note: if your goal is to report the findings to regulatory agencies (FDA, EMA), it's commonly advised to do that, please find the EMA guideline linked below. But not because of importance of the baseline differences (it's an RCT, so even if present, differences are due to chance and must be ignored), but to "filter out" these potential fake differences from the treatment and improve inference. The adjustment applies for both analysing post-values and change from baseline*.

----

* Just a warning about modelling the change-from-baseline in RCTs. It poses a serious problem if there was actual "virtual" (regardless of the randomization) difference at baseline and also the between-arm difference in post-values, of similar magnitude. In this case the slopes connecting pre-post in both arms may be close-to-parallel suggesting "no difference in differences", EVEN if the post-values DO indicate statistically signifiant (and hopefully also clinical) difference. This is called the Lord's paradox and it's a sound reason to use change scores in observational studies and post-values in RCTs. Change scores = parallelism of slopes approach actually "looks" at the baseline differences (which makes a lot of sense in non-RCTs and not in RCTs), post-values approach ignores the baseline difference and looks only at the follow-up post-treatment findings (makes a lot of sense in RCT and not non-RCT). It's up to you which approach is the best for you.

I am using a quartz cuvette.

I agree with Dr Ali. However the programming is up to the researcher IMHO. best wishes David Booth

To generate the statistical correlation matrix between the principal components (PCs) and the input image bands in ENVI, you can follow these steps:

The correlation matrix will display the statistical correlation coefficients between the PCs and the input image bands.

To complement Clemens answer, I would like to mention backward linear prediction, which is a standard procedure in almost all NMR processing software packages. It helped me a lot for measuring 11B NMR spectra using standard (non-quartz) NMR tubes.

Good luck,

Vladimir

Noor Ul Ain

I suggest you use Structural Equation Modeling (SEM), widely used for moderation analysis.

Dear Bertrand Cornu ,

thanks for your suggestions.

I would like to underline that the protocol that I used was 15' long (and it's reported to induce LTD in hippocampus).

Regarding your second suggestions: are you sure that blocking NMDAR with APV might be correct? LFS mediated LTD is a NMDAR-dependent mechanism. How can I induce LTD blocking NMDAR? Has you a reference for that?

I think that the speed of the flow rate can influence fEPSP amplitude. You may believe that your flow rate is constant between conditions, but if your system is gravity fed, it could be that the flow rate varies depending on the height of the solution.

Apostolos' idea about reference electrode is worth considering, but I believe that changes between reference (ground) and recording electrode will influence the absolute baseline values, but not the amplitude of the fEPSP.

If the sample is small, we use the non-parametric statistic, but if the sample is large, we use the parametric statistic

You should not calculate the effect size via correlation. I think the best way in this situation is to calculate the mean difference between the baseline and after-treatment for both intervention or control groups. Then, you should use this mean difference as an effect size, and due to the presence of an equal number in baseline and after-treatment groups, you can sum the standard deviation and divide 2.

If the intervention seeks an individual effect on the subject and not a group response, there is no problem with the methodological design. Even when the effect of the intervention is intended to be evaluated in aggregate. The problem arises when the group reaction to the intervention intervenes in the expected effect on each subject. Under this assumption, the intervention must be carried out simultaneously with the test group.

Hi Lokeshwaran .S ,

As a matter of fact, when you subtract the baseline value to the actual value, you do have the possibility of going below 0. This is fairly normal in several circumstances.

Let me give you an example: your stimuli are very bright in luminance, but, in the baseline, your participants watch a black screen. In this case, you will record a very high value in your baseline (say, 6 mm, that is, mydriasis). Then, as soon as the stimuli are presented, your participants' pupil dilation will normally decrease due to the high luminance (i.e., myosis). So, the average pupil size could be - as an example: 2 mm. When you subtract your baseline, you will reach the value of -4 mm. This does not mean, of course, that the pupil dilation was negative, but that it was below the baseline level. In cases like this, the baseline value should never be considered as it's unrepresentative by definition.

To avoid such a result, you have two possibilities:

1. Convert your pupil dilation data into percentage change pupil dilation. This procedure was described by Lemercier and colleagues (2014). I used it and explained it thoroughly in this paper (attached in pdf):

Ansani, A., Marini, M., D’Errico, F., and Poggi, I. (2020) How Soundtracks Shape What We See: Analyzing the Influence of Music on Visual Scenes Through Self-Assessment, Eye Tracking, and Pupillometry. Frontiers in Psychology 11:2242. doi: 10.3389/fpsyg.2020.02242

2. As a secondary choice, after the baseline subtraction, you can normalize your pupil dilation data; namely, the variable's values will range between 0 and 1.

Here's a link with the formula:

In both cases, you will lose the real data indicating the pupil dimension in mm. However, unless you're doing clinical research in the very field of ophthalmology, or similar domains, it's fairly rare to resort to pupil dilation in mm. What counts the most is the extent to which the pupils dilate in correspondence of some stimuli.

I hope I helped you ;)

Please don't hesitate to contact me for any further information.

All the best,

Alessandro

Lemercier, A., Guillot, G., Courcoux, P., Garrel, C., Baccino, T., and Schlich, P. (2014). “Pupillometry of taste: Methodological guide-from acquisition to data processing-and toolbox for MATLAB,” in Quantitative Methods for Psychology, Vol. 10, (University of Ottawa; School of Psychology), 179–195

This really isn't a question for stats folks, it depends what your model is for the mechanism that may be reasonable for some change. So, more information is necessary about this.

The question is too general. To generate the basic idea of the algorithm, you need to know the detailed statement of the problem. Typically, the process of developing an algorithm starts with the identification of your problem complexity class. If there is no evidence or clear feeling that the problem is NP-hard, then it is reasonable to try to develop a polynomial algorithm for solving it. Such algorithms are usually based on the use of specific properties of the problem. Sometimes it is possible to construct a polynomial algorithm based on the general scheme of dynamic programming, taking into account the specific properties of the problem.

If the problem is known to be NP-hard, then branch-and-bound methods, dynamic programming, and their modifications often work well for a relatively small problem dimension. Sometimes it is possible to build a successful formulation of the problem in the form of an integer programming model, followed by the use of appropriate methods or ready-made software. For high-dimensional problems, you can either use well-known metaheuristics, or develop your own approximate algorithm. In the latter case, success is usually based on the use of the problem properties. As you can see, in any case, it is useful to start by studying the specific properties of your particular problem.

Here is an interesting video that helps with that

Dr. Mwoya Byaro kindly assist

Sorry, for the long delay!

Our music excerpts are 70 seconds long and we have 20 songs altogether (10 neutral/10 positive).

Would you mind sharing this groove literature with me? I would be really interested in reading papers showing this connection between moderate music complexity and peak moments!

ad 1. Yes, we thought about that but many scientist I talked to argued, that the brain's network activity is so complex and sophisticated that one cannot (should not) simply subtract this confound from the signal... But I will look into it again! Thank you!

ad 4. Mhm this is a very good idea! I will look into that aswell.

Thank you for your responses Connor!

Dear John,

Please, read my paper and see some options:

Regards,

Gokhan

Dear Ambrin,

I am late to write, but this can be helpful for the people newly experiencing the same problem. For a solution, please read my paper's methodology:

Try to check a book on: Design and Analysis of Experiments, just be sure that you are doing the right design. Regards.

If you want measure fasting glucose, you must tak it beFore your mice get food and they not eat 8 hours. If you want measure prandial Glucose you must take 2 hours after meal. You can make it for baseline bolos glucose level

You may like this publication of the topic:

doi: 10.1096/fj.201801460R

Best, Karoly

Piyush,

Please read the methodology of my paper:

https://www.mdpi.com/2076-3417/12/3/1593

you will see what you can do.

Perhaps you have found the answer, but apparently the Helium cylinder can be contaminated with N2 which explains the high N2 ratio.

Perhaps I am being dense here, but percentage does take frequency into account,

%A = ( freqA÷total ) ×100

Am I missing something here.?

Best wishes David Booth

You can use a plain ANOVA for repeated measures (time) according to Bhogaraju Anand suggestion, as for normality test, forget it...it is not essential and parametric tests are sufficiently robust as for deviations from normality (see attached file)

Dear Gitashree Gogoi,

I suggest you to calculate carbonyl index according to the method explained in the attached article. I hope it would be useful.

I would say, if you need it for a study purpose not for exhibition (just kidding), I would suggest to check if it is possible to use the original working device with just unplugged wires in a sham group. Just and idea, good luck with your paper!

Baseline drift is a common problem in UV-vis spectroscopy - see https://ibsen.com/resources/detector-resources/subtracting-dark-spectra/ for a useful outline of the issue. I am glad that you were able to pinpoint the source of error in your experiment, Zaid Assaf. Increases in dark current (e.g. due to heating) seem to be a frequent offender, though loss of minor absorbing species (e.g. consumption of acid during a reaction) or changes in the solvent environment (e.g. in gradient chromatography) could also contribute.

My advice to William Letter is to read questions carefully before jumping to conclusions regarding the poster's knowledge and experience level. Condescending comments like "This is basic spectrophotometry" are not constructive and may deter other researchers from seeking assistance online. In this instance, you mistook the poster's issue with baseline drift as a misunderstanding of the Beer-Lambert law, despite repeated clarification. I hope that you will take greater care in your future replies.

Nice signal. The decrease is probably due to bleaching like both answers said. Continuous and more exponential than linear. Nothing dramatic you can either decrease the illumination power to reduce it or you can detrend it post hoc with your software or choice. The fact that it comes back up on the blue trace is a bit strange but probably due to the way you calculate your DeltaF/F0.

You might also try to take a look at this article:

We used a little bit more sophisticated approach than the one Olga Seminck indicated. However, I would simply try to combine both approaches.

The question is about the baseline as he has shown attached with @Ong Hui Ling question. My answer is not to use such base and I have given a suggestion above.

Also, I see that you are running at 210nm. Such a low wavelength will show just about every change in the system (normal, as 210nm is not selective for anything and will show most "changes"). Additionally, you have the software "Reference Wavelength' feature turned ON. Please turn that feature OFF and read the article below for more info why it so important to not have it ON and have professional training in HPLC operation (esp DAD) before you run samples.

Studies with bigger sample sizes should have higher weights, that's the only logical thought I can imagine here.

A few comments:

You could also try the package I am developing in R: https://github.com/ethanbass/chromatographR. It doesn't have a parser for chromeleon files, but I could probably write one for you if it's really just a text file?

As it appears you are using HPLC for the first time and it takes many years of professional experience just to achieve a basic level of experience with HPLC, please contact someone local at your school to help you. As a student, you should not being doing this alone. Having an experienced chromatographer to help you on-site setup a method and run the analysis will help you be successful. The manual that came with your HPLC column also has a lot of useful information regarding the use, cleaning/washing/regeneration/storage too (do not store it in water !!!!) . Be sure to contact Bio-rad if you have column questions.

Also, the fact that your column temperature is set to 80C and your RID at 55C, will result in a slow temperature gradient forming. This may cause long term baseline drift. Since temperature is used as variable in method development (for your specific application) and temperature changes take a VERY LONG TIME to equilibrate, on column, be sure to insulate all capillary lines leading from the column to the detector (see article linked below) and please allow for LONG equilibration times.

Hi Chris

I am facing similar issues of baseline stability and varying retention times. what I understand is RID is highly sensitive to temp and it takes a good amount of time to stabilise the baseline. I am operating at 40 degress both column oven and detector.

"What test would offer insight as to group x condition?"

Only a parametric model can do this. As soon as use ranks (i.e. some kind of "non-parametric analysis") an interaction is not meaningfully interpretable.

There are more important things to consider when analysing an interaction: it makes a difference if you assume that effects are additive or multiplicative. A meaningful interpration also requires that that the observed interaction is not due to ceiling or floor effects.

It's easy to do "some test" and to get "some result", but it is tricky to get a meaningful interpretation. I suggest to collaborate with a statistician.

Hi, Nikolas Rizzi

Given that the EEG signal is non-stationary, I'd say that it is perfectly possible for the signal to 'drift' across trials. In fact, I have at some point recorded data with such a behaviour. In any case, there are drifts that can be removed with the appropriate type of filtering (see, for example, this link: https://benediktehinger.de/blog/science/electrode-drift-in-eeg/). I would recommend normalizing your task data by the power at baseline, so that, if such a drift is indeed present, you can consider it in both task and baseline.

You could also always compute the baseline power and check if it, indeed, increases over time between your trials.

Does that make sense?

Dear Fikret Ustaoğlu

"The objective of BASELINE is to publish short communications on different aspects of pollution of the marine environment. Only those

papers which clearly identify the quality of the data will be considered for publication. Contributors to Baseline should refer to

‘Baseline—The New Format and Content’ (Mar. Pollut. Bull. 60, 1–2)." ( ).

Typically, a drop in absorbance is due to the change in RI or UV spectrum particularly between the mobile phase and extraction solvent and is apparent in the solvent front. This is different. I would ignore it because your sensitivity is set too high. I would be concerned if it was ~0.1 mAU or 10X higher than what you have in your chromatogram. Your molecule of interest will be at least 100 mAU or 50,000-100,000 peak area units.

Jane Mbijiwe Because you are working with the same population and the same treatment/intervention, try utilizing the Wilcoxon Rank Sum Test. It will assist you in getting to know the group that benefited the most from the intervention both before and after.

David J Walker is it possible to cite the publication in which you found informations about normalisation please

Thanks

Hi Lauren, for me, for some reason, LTP is inducible only in like 20% of slices at 50% of amplitude at first PS response. But if instead, I take ~60-70% of this amplitude, the success rate increases to 75% (with no PS at during LTP recording).

And also I think it is better to deepen the stim much deeper (like penetrate it >half of a slice), so more fibers would be stimulated. But I didn't try it yet.

By the way, since I'm answering pretty late and have similar problems (wish to have 100% success rate at 50% stim of max), could you please tell, did you solve the issue? If so, how?

Thank you Joao and Nik for your replies.

I was able to baseline correct from the command line/Matlab space, as you suggested Nik. I used the following code, which saves a new dataset with the applied baseline correction:

pop_saveset(pop_rmbase(EEG, [-200, 0]), 'filepath', 'C:\Users\s15aa\Downloads\eeglab_current\eeglab2021.1\EEG Epochs and Baseline Correction\PSVTR08_hard_baseline_corrected')

Hi dear Sara Guemar,

The first and most important thing is that lars works as a stochastic tool. Therefore, it is important that you understand how the LARS works.

Firstly, I want to emphasize a key point. You mentioned that you chose 2006-2010 as your baseline. If you are going to publish your research as a scientific article, you should choose the standard period of 1980-2010. Lars is calibrated for the period.

Secondly, you asked what is the start of the period selected for 100 years? After choosing a GCM model and scenario, you must select a period. When the selected period is, for example, 2041-2060, it is not important whether you choose 100 years or 10 years; the model will generate data for that period. My description of it is as follows.

As I told you lars is a stochastic weather generator. It means the output of Lars is not a time series. This means that the numbers for 2042 are not after 2041. Thus, averages (or std) of series can only be used for analysis. LARS is not a prediction tool, but a projection tool.

You can test this out for yourself. For the period 2041-2060, generate data once for 20 years and once for 100 years. Then compare the average of the temperature and precipitation for the two series. You will see that the means are very close to a very good approximation.

Finally, you should know that you are allowed to analyze the results on a monthly basis since Lars does not support daily changes unless you manually create scenarios based on daily GCMs.

To get a comprehensive understanding, first an idea about baseline characteristics is compulsory. These characteristics are the medical, demographic and prognostic variables in Clinical trials. When an error occurs in creating intervention groups in baseline characteristics when participants are non corresponding to prognosis, this is called Baseline imbalance.

Hope that it helps.

Dear Lisandra Almeida,

If I would be your help, my email: miregech897@gmail.com.

Good luck!

Paired sample t-tests could be a simple approach to look at differences in a single variable from pre treatment to post treatment.

OK, I assume you have baseline and endpoint measurements for each of the 239 subjects. I was lazy and used the Comprehensive Meta-Analysis (CMA) software to do the calculations for me. There are two possible ways you might calculate the standardized mean difference (SMD). The first one (see Test1.jpg below) is probably the best procedure. It makes use of your paired t-test p-value. You do not have to make assumptions about the correlation between the baseline and endpoint values like I used in Test2.jpg. In Test2.jpg (rows 2 and 3), I used assumptions of r=0.9 and r=0. These are extreme correlations but it is nice to see they are similar to the SMD calculated using the p value (i.e., that in Test1.jpg). If you want a reference for the CMA software, use the book by Michael Borenstein et al. (i.e., Introduction to Meta-Analysis, Publisher: Wiley [2nd edition], 2021). SD of Difference = SE x square root of 239