Bridging IT and Statistics: A Guide for Statisticians

[00:00:00] Alexander: Welcome to another episode of The Effective Statistician. Today, I’m super excited to have Pablo with me here, a statistician that turned into a programmer and that has a lot of interest in the relationship [00:00:20] between biostatistics and IT. And just from my personal Experience, that interface is not always easy.

[00:00:33] So, but before we dive into this Pablo, maybe you can introduce yourself shortly. 

[00:00:39] Pablo: Thank [00:00:40] you very much, Alexander, for having me here as an effective statistician. And to be a part of one of your podcast editions. Well, I’ve been in the pharmaceutical industry for more than 40 years. I started as a [00:01:00] statistician and held several positions also in IT, marketing sales, came back to statistics, and then I discovered my passion, which is statistical programming. That’s about me. 

[00:01:15] Alexander: Awesome. So, in terms of your experience [00:01:20] with the biostatistics function, and of course that includes programming, how is the, your experience in terms of the relationship between IT and all functions? 

[00:01:33] Pablo: Yeah, that’s a very interesting question, Alexander. Thank you for making it.

[00:01:38] I tell you, when I [00:01:40] started, I started on paper. And I had a very big paper, a huge paper, where I had my rows and my columns. It was my database. And well, I kept statistical formulas in my head and looked at documentation also if I needed it, but books [00:02:00] books, not, not as these days.

[00:02:01] It’s beautiful these days. And then all this, all this started evolving. From well, mainframes, of course, that existed, and then standalone PCs client server solutions. And these [00:02:20] days, the buzzword besides AI is the cloud. 

[00:02:25] Alexander: Yeah, everything is in the cloud. Yeah, yeah, yeah. So and of course, IT enables a lot of all the things that we do.

[00:02:34] Yeah, without IT, we, you know, couldn’t do lots of the more modern [00:02:40] day calculations like Bayesian tools. That’s brilliant. really just made possible only after very much increasing the calculation power. So in my early days the Bayesian statisticians always said like, yeah, if we would be able to compute it faster, then we could do these kinds of [00:03:00] things.

[00:03:01] Pablo: Yeah, I can imagine. I can imagine that. Well, when, when we started using You were happy if you could do a PROC FREQ, or a PROC REPORT, or a PROC TABULATE. Now this has evolved a lot. You have SAS. You have SAS [00:03:20] Foundation, Viya, you have R, you have Python, and those are you have SLC, that is a SAS analog, analog software.

[00:03:33] So you have a very broad spectrum of tools these days that [00:03:40] allow you to do these very complex tasks. statistical analyses, and develop the models, and have very fast results, of course, depending on your hardware resources, depending on your on your network resources but, but that’s it these days.

[00:03:59] You [00:04:00] can solve very complex topics in a more easy and comfortable way than years ago. Fully agree with you, Alexander. 

[00:04:08] Alexander: Yeah, and as you speak about hardware, about speed, about cloud, these services all are provided [00:04:20] through our IT departments, either internally or externally. Now what are actually the more specific services that we as a power statistics community rely on so that we can do our day to day work?

[00:04:36] Pablo: On the one side, you have the hardware. [00:04:40] And on the hardware side, you have, you have, you could have a stand alone solution where you have the software as a SIG client. The SIG client is where the software runs on a PC like Microsoft Word. 

[00:04:51] Standalone SAS with display manager, which is no longer being used as such in, in, in, let’s [00:05:00] say, in, in client server or cloud solutions.

[00:05:04] You can talk about SICK clients. You can talk about SICK clients. When we think about the studio and you can, if you are an enterprise guide. SLC Workbench, RStudio for example. [00:05:20] 

[00:05:20] Alexander: So, so let’s first start with that. So the kind of standalone that is when everything is on your laptop, PC or whatsoever. Yeah. And it all stays there.

[00:05:32] You don’t need an internet connection. Everything is there, more or less, yeah, and then the client [00:05:40] is basically like in the former days where basically your entry, your laptop or whatsoever is just the interface to the computing environment that sits somewhere else. And that could be kind of in the cloud, or it could be in a server that is related [00:06:00] within the company, so to say.

[00:06:02] Pablo: Yes, yes, that’s completely, completely true. Yeah. Point is, you have the server, it is called, the solution is called on premise. You have a server in a company, maybe in your own country or in some other country. [00:06:20] And that is where you have the core software. 

[00:06:24] Alexander: If you talk about on premise, that is basically on premise is kind of the opposite.

[00:06:32] Of in the cloud, so to say, so, so on premise means you own these kinds of servers or you have rented these, [00:06:40] but they, they are kind of, you control them, so to say. 

[00:06:43] Pablo: You own the service, you own the service. And you can also rent them. If you want a software solution outside of your company, you can rent the server and there, let’s say, let’s put it like this.

[00:06:57] You have their SAS. The, [00:07:00] the core SAS, all the elements that make SAS happen, and we have the client on the other side. I call it a sync client in this case, because in the case of SAS Studio, all you have to do is have a link and through that link access sAS. It is [00:07:20] also an IDE, right? It’s also an IDE, since you can write code. You can write code. Push the code to the, to the server, to the cloud, and then it comes back. 

[00:07:34] Alexander: Okay. 

[00:07:35] Pablo: And yeah, and that is something difficult to [00:07:40] understand for, for common for the common user, for the, for let’s say for the statistician, also many times for the statistical programmer, that separation of the, of the server part or on premise or the, or the cloud part, where.

[00:07:56] Where your core software relies on and [00:08:00] the thin client part where the the software used to To connect or interact with the core software and send information and retrieve information in the form of results, for example. 

[00:08:17] Alexander: Okay. So one of [00:08:20] the things that your IT department should kind of help you with is to decide Whether you want to have a local solution, a solution on premise, or a solution in the cloud.

[00:08:35] Pablo: There are several departments that interact there, mainly IT, [00:08:40] of course. But a business owner, let’s say a very proficient SAS user who could be the business owner of, of, of, of, of SAS would also support the best decision. And of course you have your enterprise, let’s say your finance manager who says, [00:09:00] okay, we can do it, but look at the costs.

[00:09:02] So it’s an integration of many elements, but mainly of course IT, the business, the business owner, finance, and of course the company where you’re going to purchase the software from, or where you’re going to get [00:09:20] the license from for a specific time period. Let’s say SAS, let’s say Altea. In the case in the case of, of R and R, R on the cloud or R studio, that’s a different story, but there are some other issues to consider.[00:09:40] 

[00:09:40] Alexander: Okay. So that is kind of part of the hardware that we need. Are there any other important hardware elements that we rely on as statisticians? 

[00:09:52] Pablo: Yes. Yes. If you are on a client, you are, you, you are using your notebook or your PC as a client, you should [00:10:00] have an amount of, of memory, I would say at least 16, 16 gigabytes of raw memory.

[00:10:07] 500 mega gigabytes of, of hard drive, and yeah, a user graphic card. That’s so that the process runs smoothly, that’s just the [00:10:20] physical part, and then what is, what is very important is the connectivity, the network connectivity. So if you’re working from home or you are in your office, in the office it’s easier because you are directly connected to the, to the source, to the server.

[00:10:38] If it relies in the same [00:10:40] country, if it relies outside the company, where like he takes care that the connection is stable. But in the case of a user working from home, you have a download and upload process. Download is when you send the information, upload is when the information comes, comes back to you.

[00:10:59] [00:11:00] And I would recommend having at least 100 Mbits. For download and 20 mbits for upload minimal. That was my recommendation on the one side, on the other side comes topics like VPN firewalls on the IT side, as [00:11:20] stability of the server have the server big enough to, to support the software, have the server for your, for your data, for your programs all it has to be taken into consideration.

[00:11:32] And the stable, and the stable communication line. That’s a challenge these days. It’s beautiful, but you have to [00:11:40] take several considerations too, so it works smoothly and properly. 

[00:11:44] Alexander: Yeah, so these are kind of all the security aspects, isn’t it? VPN, 

[00:11:50] Pablo: Yes, firewall, VPN. antivirus software, spyware software, all those are security topics. Yeah. 

[00:11:59] Alexander: [00:12:00] Yeah. That is definitely another service that we very heavily rely on. And of course, given that we work directly with highly sensitive data and pretty costly data also it is really important to make sure that everything from a security side of things work well. 

[00:12:19] Pablo: That’s a [00:12:20] must, especially as you say in our environment, the pharmaceutical industry, where everything is very regulated. The, you are owner of the data, the data are very sensible, that has to be taken care of. You can of course play around, I just read, it was a couple of days ago something about [00:12:40] JetGPT now for data analysis, you can integrate it with, with very well with Google Drive and do fantastic things, that’s beautiful but the point is security validation, checking of, checking of the data, well, checking of the data, you have to do by any means, but the main issue there is [00:13:00] security.

[00:13:01] security of your data, yeah, and that your processes, your analysis processes run well. 

[00:13:08] Alexander: Yeah, that is another kind of interesting area. When, when we talk about data manipulation programming storing, these kind of things all these [00:13:20] Documents need to sit somewhere and so this needs to be in in a regulated environment and a validated environment as far as I understand.

[00:13:31] So how does IT help us in that regard? 

[00:13:37] Pablo: Well, then there comes another component in a [00:13:40] company. There comes now IT the quality control department, who takes care, for example, and gives the guidelines on how you should do your your your validation. Of course, with the help of IT, and let’s put it as I know it, with the support of SAS.

[00:13:57] SAS, for example, has an [00:14:00] IQ OQ process. Besides that, you have to do your own, prepare your own programs to, to, to validate that when you migrate from, from one SAS version to a newer one, that everything continues running, running fine. Yeah, that’s very important. And you should stack in any case, [00:14:20] be it SAS, that, that would be my, my, my recommendation, my choice that you have a ver, also a versioning component.

[00:14:28] Let’s say the latest SAS version of SAS foundation, SAS analytics, is That’s 9. 4 M8 [00:14:40] M9 is coming this year, but you have to keep track of that. Yeah, and also the same client component now this year enterprise guide 8. 4 is coming out Of course you don’t, if you don’t have the nine, the the buyer version, you can say you, you key stack on enterprise K eight four and H eight three.

[00:14:58] Pardon? Not, not, [00:15:00] not, not eight, four, and because you don’t need the rest. But point is you have to have a clear concept of your versioning. And that also is is the same for, for r. In I web to stack on a spec on specific versions of r mainly of R which is the core and also of our [00:15:20] studio.

[00:15:20] Yeah, that’s very important. 

[00:15:23] Alexander: You, you just mentioned two acronyms, IQ and I think OQ what does that mean?

[00:15:32] Pablo: IQ is the installation qualification tool that ensures that SAS software [00:15:40] is installed properly according to the manufacturer’s specification. In this case, SAS. And that environment is set up correctly to support the software. 

[00:15:52] Alexander: Okay, and that is basically the same for any software, isn’t it?

[00:15:56] Pablo: Yes, and that [00:16:00] generates a document that you store for auditing, for example, and everything should be okay.