WARNING LONG POST!! – I will try to keep posts much shorter from here on out!
Up to this point, I have not been keeping track of the actual number of applications I have sent out since my graduation. It is easily over a few hundred, as I was sending out at least 10 per week right after graduation. The rate of application declined as we approached the holiday season and no one seemed to be hiring again. However, with the new year in full swing, job posts seem to be back to normal. Not that I expect a following of any sort to give me feedback, I’ll still post it here.
What sort of work am I hoping to obtain? Something that uses my statistical training from reasearch methods using excel and SPSS software would be ideal. This is much more realistic than looking for someone to hire me based on my M.S in psychology unless I was to go into mental health. My degree did not have a single abnormal psychology course in it, so that would not be a realistic avenue for me to pursue. I always get the feeling that potential employers see my resume and are not entirely sure what to do with me. They see that I have an M.S. in psychology and that seems to be a strange thing in and of itself. I feel like a lot of people do not realize that there is a difference between an M.S. and an M.A., actually, so let’s have a quick chat about that!
M.S. = Master of Science. These degrees typically require that you work in the research labs, take research methodology classes as a requirement as well as at least a few semesters of statistics.
M.A. = Master of Arts. These degrees are more directed towards taking more courses and learning about theory, history, and sometimes practice.
The major difference between these degree plans is the requirement of lab work. But what exactly does 2 years as a research assistant do for you? I suppose we can look the typical workflow for a project and you can see how an assistant gains experience.
During the first meeting of the year, we discussed all the major projects and which ones need more support. I was informed that one doctoral student needed help running participants through his study and that I should work with him on his project. At this point, I would have to begin reading up on the method they are using and the theoretical justification for their experiment. Typically, I would be referred to a few supporting publications so I have a clear understanding of the project. Next, I would the experimental task myself, or at least a part of it, so I can be more informed when training a participant. This is still pretty simple in most cases. For my project, people were looking at faces on a screen and hitting buttons to indicate if they had seen a face earlier in the test or not (an eyewitness test). Fairly straight-forward.
Next, I began running participants using online scheduling systems. This was a fairly simple task as well. I would place my available times up on the site and participants would indicate that they would come at a particular time. After running a bulk of the participants we began putting the data into excel from E-prime presentation studio. Things got a bit tricky here but it only took a few hours. You have to filter out a lot of the excess information that E-prime produces so you can find your values that are critical to the experiment (such as what did the participant answer, but ignore how long it took them to make each response). In other designs latency (response time) is quite important, but we did not have a hypothesis that took latency into account, thus it was not placed in our database. No worries, though! Should we need that latency data later down the line, it is all still there in the raw format, just not in the cleaned databases!
After this we placed everything into a really massive set of excel tables. Each participant had over 100 responses that had to be coded in. This was done through formulas in excel, but getting the spread sheet formatted in this manner was a frustrating task. I joined the project when it was running experiment number 3, so most of the work on excel was basically complete. After reviewing the Excel documents I notice some really large errors in the scoring of participants that seemed related to their subject number, and not their actual ability. Why would subject number 20 of experiments 1, 2, and 3 always do really bad? Not just bad, but almost like they were doing this poorly intentionally. Keep in mind that subject number 20 was different for each version of the experiment! It seems highly unlikely that this was random, and we had to start by looking at our excel formulas. We were hoping this was merely an incorrect cell reference that could be solved with a single edit. Well, after regressing things back to that point, we came up empty and had to begin combing through the E-prime coding. This is much harder as I have less experience with it. After a few hours of frantic searching, we found that answers for subject 20 were coding in backwards. Nothing is worse than having a problem with the data and being unsure where it is coming from, but we found it and solved it soon enough. Later down the line, it also turned out that a number of coding errors were committed in Excel that appeared in Experiment 1, 2, and 3. I was brought on board in experiment 3, so I tried to work with the original author on finding out why he made these errors. We cleaned it up within a week, having to manually edit a large number of entries, but it all turned out clean. That’s as dramatic as things tend to get in the lab!
So now that we have all our values in Excel and the data set has been cleaned of all imperfections, I began to put the values into SPSS. We can use Excel to find simple things like correlations, but it is harder to use it to make predictive models, like a statistical regression. A statistical regression seem similar to a correlation, but instead of comparing one variables to another, you are comparing a set of variables to one dependent variable. Another key difference of regression is that it infers causality. So your predictor variables that you choose are supposed to be causing the changes in the dependent variable. This must be supported through the literature for you to just decide something needs to be a regression, otherwise it’s just data mining. Anyways, we put our set of predictor variables in and we did not find much of any significance for experiment three. (I will explain regression in excessive detail in a later post, as this one is already getting too long!)
Now, I’m leaving out a large portion of the details of the experiment as it is still a work in progress. I independently ran experiment 4, and I was able to add personality variables to this test and found some startling results. This project turned into my thesis after I spent enough time with it, but that’s for another day.
So as a research assistant I took a project from running participants, data entry, data cleaning, all the way until the statistical analysis. When I did this for my thesis project, I started from the conception of the project all the way to the final presentation of the findings in APA format. I think the final draft of the thesis poster was draft 11… or was it 14? I can’t remember. I just recall looking at all the text and it just becomes a blur after having looked it over 1000 times, haha!
In closing, my time as a research assistant was extremely hands-on. I rarely sat back and watched people tell me how things should be done. They typically instructed me one time, and then had me try it out with supervision, and if it went well, I would do it on my own from then on. It was a great learning experience and I honestly hope to put this skill set to use again in the future.