Research questionnaire / 4

I finished reading the monograph which I mentioned the other day, "Survey questionnaires: handcrafting the standardized questionnaire". After all, it is only 88 pages long. Whilst there is some interesting material within, it wasn't as helpful as I thought it would be. There are two major problems with the monograph which render it of little use

1. It assumes that the questions are being asked of the general public and not of a specific population. That said, it's interesting to note that many of my respondents who use Priority, an ERP program, didn't know what ERP was.
2. It assumes that the questionnaire will be used in a face to face or phone interview. It was written in 1986 before personal computers became truly widespread. Thus there is concern that the respondent may not have heard the question fully and may have substituted homonyms (the example given is hearing "prophet" instead of "profit"). Such problems are simply not going to occur with a computerised questionnaire.

I also received feedback from my mentor regarding ethics and why there was no material in the IBR courses about constructing questionnaires. Here is part of his answer: it is simply not practical at the IBR stage which has to give generic background to research. It does not and cannot cover all the methods and techniques that DBAs could draw on; if it did you would have to learn an immense amount of material that you would never use. After that, it would all have to be customised to your needs. It is much more efficient for you to choose a la carte rather than having to consume the whole menu. Your own questionnaire will be a survey type questionnaire and few of our DBAs do this. Much more efficient for you to develop this under the guidance of an experienced supervisor. 

I find the comment "few of our DBAs do this" (ie create questionnaires) intriguing. It implies that most of the theses are what are termed "longitudinal studies", where the research data is obtained from a small population over a long period of time - by structured interviews.

Research questionnaire / 3

In light of some of the responses which I have received to the questionnaire which I sent to all email users of my company last week, I've been considering whether creating a questionnaire is something anybody can do or whether it takes special skills. Certainly, there was no mention of creating questionnaires in the DBA courses; maybe it is assumed that the researcher uses a questionnaire that already exists.

Some of the papers which I have read include a questionnaire, but normally the questions are either too widely based or out of date for my purposes. There is one paper - actually a non-academic survey - which presents answers without the questions, and of course it seems that this questionnaire might be the most useful of all. I wrote to the contact address of the survey today; the contact person replied very quickly and said that he would try and find a copy of the questionnaire.

I then had a look at the Academia stack exchange site to see what they had to say about questionnaires. The first question which I found on the subject had answers which strongly advised that creating questionnaires is a skill which must be taught. One person wrote about an 80 hour course!

Another answer mentioned a monograph called "Survey questions: handcrafting the standardized questionnaire" by Jean M. Converse and Stanley Presser. I was able to find an online copy of this short book (88 pages) and have started perusing it. On page 10, the following is written: "Because questionnaires are usually written by educated persons who have a special interest in and understanding of the topic of their inquiry, and because these people usually consult with other educated and concerned persons, it is much more common for questionnaires to be overwritten, overcomplicated and too demanding of the respondent than they are to be simple-minded, superficial and not demanding enough".

I very much understand that paragraph: my questionnaire used several terms which were obvious to me but not to most of the respondents.

I shall ask my mentor two questions

1. Is it ethical to use questions which appear in other people's questionnaires? I appreciate that if I use the 12-item instrument to measure End User Computing Satisfaction (EUCS) developed by Doll and Torkzadeh (1988) in full, then I have to cite them, but what should I do if I only use one question of theirs, along with one question from someone else?
2. Why is there no material in the DBA courses about questionnaire development?

DBA mentoring period commenced

After another few weeks of having been left in limbo as regarding the DBA, I was informed last night that I have paid my mentoring fee and that Professor X has now been appointed my mentor.

I think that I sent four times the fax with my credit card details with regard to the mentoring fee; I can only hope that I am charged only once!

I wrote a note to the mentor last night, inclosing the draft of the research proposal as it currently stands. I am to expect a response within a few days. Hopefully I can now start moving purposefully instead of floundering a little.

[SO: 3417; 2,12,30]
[MPP: 356; 0,0,4]

Boy, was I wrong - programming naivety

It's a well-known fact in computer science that the higher the programming language, the less efficient it is. This is why programs written in assembly language are smaller and faster than those written in a high level language. But also within a high level language, there can be low and high level constructs: the low level construct may require more statements from the programmer but will run faster (because the simple, high level construct hides from a user a multitude of low level constructs).

Let's say that there is a table whose tuples are composed of the following fields

ID - longint (four bytes)

name - varchar (thirty two bytes)

other data - sixteen bytes

comments - varchar (128 bytes)

Each tuple will thus comprise 180 bytes, assuming that there is no overhead. Let's say that there are one hundred tuples in the database -  this is very small, but suitable for a 'master' type table; such a table will probably have more fields but no more tuples. The scenario is that the entire list of tuples is displayed for the user so that she can choose records for updating.

The truly naive programmer will issue a database command - for example "select * from table" - in order to retrieve all the tuples from the database. Result: 180 * 100 = 18,000 bytes are retrieved from the database; in reality, the amount of data will be less as the 'comments' field will not be completely full in every tuple. A more realistic figure would be 116 bytes per tuple, or 11,600 bytes for the table. If the database sits on a network, those 11,600 bytes have to travel along the wires. The truly naive programmer will not be aware - or will ignore - the time necessary for the data to arrive. All he knows is that he can get them whenever he needs, using one small but powerful command, 'open'.

I learned at least fifteen years ago a style of programming with databases which is fairly efficient: this separates the displaying of the table from the editing. For the display, I need only the id and name - thirty six bytes per tuple - so I would request from the database only 3,600 bytes: a third less or three times quicker. The tuple to be edited will be retrieved in its entirety.

Yesterday, I had an epiphany which shows just how naive I am too (or as I prefer to call it, "blinkered"). The epiphany connects to the opening paragraph of this blog and especially to the that one small but powerful command, 'open'. When one opens (or reopens) a database query, all the data has to be retrieved from the database.

What if one is deleting a tuple from the table? I would issue a 'delete' statement with sql which is as efficient as possible, but then I would issue an 'open' command in order to update the display. Why am I requesting all the data which I already have when all I need to do is remove the deleted record???

I use the clientdataset component in Delphi, which is extremely powerful - but apparently only partially understood. This component holds in its buffer all the data retrieved from the database, making various actions, such as on-line sorting, very fast. With regard to the tuple deletion, all I need to do is delete the tuple from the clientdataset - there's no need to retrieve from the database data which I already have in memory!

The extension of this idea hit me only when I was in the shower prior to going to bed (I do some of my best thinking in the shower): when adding or editing a tuple, there is no need to retrieve all the prior existing records; it's enough to retrieve only that tuple then add it manually to the clientdataset. True, this will require more work from me than simply writing 'open' - but I only need do this once, and the user will benefit from this every time a record is added to the table.

The same idea applies to polling - there are applications which I have written which poll (ie access) a database table every five or ten minutes, always clearing the data held in the clientdataset's buffer then retrieving the new data. What if no changes had been made to that table in the intervening five minutes? My original idea was to check the number of records in the table and update only if the number has changed, but now it occurs to me that I only need retrieve records which have been added since the last access and manually add them. This way, I will greatly reduce the amount of network traffic.

Of course, all of the above is correct if only one person is updating the entire database. If there are three or four users (I am thinking of the occupational psychologist's offices), then all the users may have the same table displayed on their screen; in the current scenario, if two people add (or edit) records, then they will see the changes made by the other user (as the data is always reloaded in entirety), whereas the two users who don't make changes won't see updated data - unless they poll. In Priority - and any industrial strength database program - there is an unwritten contract that once data is displayed on the screen, the display will not change, even though there may be changes to the underlying data. So, in a sense, I had been spoiling the OP's workers, at the cost of network traffic. That era has now come to an end.

Research questionnaire / 2

Several months ago, I wrote about my initial thoughts on the research questionnaire which will be the main instrument through which my doctoral research will be conducted. In the intervening period, I've changed some of my ideas regarding this questionnaire: in the same way that I don't want to include 40 questions about cognitive style, I also don't want to include 10-15 questions which try to elucidate how well the respondent knows Excel.

I spent my 'doctoral hours' on Wednesday going over the questions which I had already established. Six or seven of the twenty five were concerned with the company's attitude to ERP; as I have already decided that I need one questionnaire for the company and one questionnaire for users, these questions were removed. I then added a few questions about training with ERP and ownership, along with some questions about spreadsheet usage and training. 

I translated the questions and answers in Hebrew then entered them into a questionnaire form in Priority. This questionnaire module is meant for performing customer approval feedback, but I could adapt the mechanism. Afterwards, I saw that this module is not that suitable for my needs: it doesn't allow the possibility of skipping questions based on certain answers: if the respondent answers that she doesn't use spreadsheets, then there's no need to ask many questions about spreadsheets!

I asked permission from my company's CEO and CFO to distribute the questionnaire amongst the company's users in order to get feedback. I stressed that I was distributing the questionnaire primarily to get feedback as to whether the questions could be understood and whether there were sufficient answers. I also stated that I was not too interested in the actual answers and that in any case the answers would not be used in the final research in order to prevent any form of bias. I asked the respondents to print the questionnaires then fill them in; I would physically collect them later (ie no need to scan them then send the scans to me by email). Approval was swiftly forthcoming.

Next morning I received five questionnaires: two from the site where I work and three from other branches. The first conclusion which I made was that people weren't reading the instructions: I specifically asked not to send me via email! I wrote a second letter to everyone thanking those who had already responded but asking the others not to scan their answers. There was misunderstanding regarding the conditional questions: one or two said that they didn't create reports intended for other people then answered questions referring to those reports.

Yesterday evening I updated my own programs with the questionnaire: as with programs which I have written for the occupational psychologist, there are three programs in the suite. One program manages the questionnaire itself, one is the exam and one is the program which manages the results. The latter is probably the simplest and hasn't yet been written. When entering the completed questionnaires into the exam program (as I were sitting the exam), I discovered several problems with the program and with the questionnaire itself (there was one conditional question which I hadn't noticed). I fixed each problem as it arose then checked to make sure that the exam behaved as it should.

Obviously it's too early to make any real conclusions about the questionnaire, but there are a few points which are already clear

• the question about previous ERP experience is unclear
• people who work close to me have received more than enough training in Priority whereas people in distant branches say that they haven't received enough
• no one yet has learnt Excel in a formal manner: everyone has learnt from a friend

I'm hoping that I will get 40 filled questionnaires; so far only one of the five that I have received made any comments, so it might well be that this trial will not help me as much as I had hoped in improving the questionnaire.

Cognitive styles and sense of ownership

My work day is rarely structured: true, there are days which are full of meetings (normally once or twice a week when I travel north) and at the beginning of the month there are quality reports that I prepare, but most days I go to work without knowing what is going to occupy my time. This flexibility allows fast user response, although sometimes it happens that three people need my time and attention simultaneously which is when I have to implement a queuing algorithm or switch into time-sharing mode. Despite the previous sentence, I am not a computer.

Yesterday was a good example: after sending my daily reports (15 minutes), I began going through the mail that had built up in the few days of my absence. This kept me occupied for about an hour and a half, then I had to solve a problem regarding undefined import costs for a part, which led me to check how many other parts also had undefined import costs. So on and so forth.

After about 11am (I start work at 7am), I went into on-call mode, waiting for someone to require my services. In between, I resumed work on the literature review for my doctorate, which has been languishing in the past few weeks. I had found a paper a few days ago which had listed cognitive style (CS) as a factor involved in users' uptake of computer programs (the paper was written nearly thirty years ago, so it mentioned Decision Support Systems instead of ERP), and I wanted to investigate this subject.

After finding several papers discussing CS and computer programs, I decided to add a section in the literature survey but not include CS in my research. This is because there are several methods to measure CS (for example, one involving 40 questions, one involving 80 questions), meaning that there is no universally defined CS. There are papers which cast doubt on the methodology of the CS exams. But for me, the main reason why I decided not to include CS as a measured factor is that the questionnaire which will form the basis of the research currently has about 25 questions - and I don't want to add another 40 questions about cognitive style!

After thinking about this for a few hours, I hit on another factor which wasn't mentioned in the old paper - sense of ownership. This seemed to be a more promising idea: there is plenty of literature about ownership and ERP (although not in the niche which I have found), and one's sense of ownership can be determined in one or two simple questions. The premise is, of course, that the sense of ownership correlates with the user's desire to create her report as quickly and as accurately as possible.

Tales of us

One website which I frequently access is the online version of 'Sound on Sound' magazine, the self-styled "world's best recording technology magazine". Whilst most of the material there is of little interest to me, there are always one or two nuggets which I read avidly. The most recent issues of the magazine are subscription only, but after a few months, all the articles can be accessed without cost.

The articles which most attract me are in the 'classic tracks' series. In 10/13, there was a very informative article about the recording of the first Clash album, which doubtless would mean more to me if I had ever listened to the Clash. In 11/13, the first (truncated) article which caught my eye was an interview with musician Will Gregory who explains how Goldfrapp's latest album, Tales Of Us, differs from its predecessor, Head First. "We wanted to step away from that and towards something more acoustic.”

This intrigued me. The name 'Goldfrapp' had crossed my radar previously as possibly an act that I might like, but whatever I heard didn't find favour in this corner. Something more acoustic, however, stood a good chance of being liked. I downloaded the album a few days ago, started listening to it in the evening and by the third track, I realised that I was listening to something really good.

I haven't listened to anything else for the past few days! I have ordered the cd, thus ensuring that Goldfrapp get their royalty from me.

The main attractions are the whispery vocals of Alison Goldfrapp (I have no idea what she's singing about, and to be honest, I don't really care) and the orchestral arrangements of Will Gregory. Some of the songs are relatively simple whereas others are very dramatic; unusually for me, I haven't yet properly elucidated the structure of the songs - I let the entire album wash over me.

I can hear echoes of Kate Bush, whose influence is strong on one or two songs, and even of Robin Frederick.

As the modern idiom has it, this record ticks my boxes.

I can hardly wait for April, when the entire article will become available for reading.

A comparison between King Crimson and Van der Graaf Generator

Following is an interchange of letters on the Hammill mailing list -

Opening statement

I read somewhere that VdGG were called King Crimson's "little brother band" and i disagree totally. THEY SOUND NOTHING ALIKE and VdGG is far better than Crimso ever were.

To which someone replied

I think this is a wrong way to put it, that one of them is better. Both are highly idiosyncratic and with a strong intellectual emphasis (remember Fripp's parallel to head, heart and hips in music). However Crimson's music is more structured, composed, 'written' if you like, more impersonal, less chaotic than Van der Graaf. Academically inclined composers tend to appreciate Crimson for that reason. (F.ex in the liner notes of "Flux" (ECM 1999) by Erkki-Sven Tüür one can read: "Erkki-Sven Tüür doesn't disavow his background and will turn to the CD player and listen to King Crimson".) Thus Crimson is much closer to the new academical music. Its hard to imagine about most of the VDGG's music being played by someone else than PH & Co. It's too expressive, personal, verbal. Too much here and now. VDGG & PH seem to prioritize the message, getting the spirit through, not so much instrument playing skills, singing, "right" sound nor compositional perfection. This, of course, is a rather schematic way to put it.

Here's my answer

This is a very interesting subject: I've been listening to both bands for over 40 years (although I didn't listen much to either during the '80s) and have often thought about comparisons. I'm only going to compare the 69-78 VdGG and 69-74 KC. Following are some conclusions -

Similarities:

1. Both had an early, "light", album (Aerosol gray machine, GG&F), neither of which hinted at what was to come.
2. After that early album, both added a saxophonist (Jackson, McDonald) which both changed the sound and made it much deeper (some might say "heavier").
3. Both were capable of playing first a heavy song followed by a "pretty" ballad (VdGG: Darkness/Refugees, Killer/House, La Rossa/My room; KC: 21stCSM/I talk to the wind, Pictures of a city/Cadence and Cascade, Cirkus/Lady of dancing waters, Lark's tongue/Book of Saturdays).
4. Both bands eschewed the blues
5. Both showed a propensity for songs with multiple time signatures (Lark's tongue, Scorched earth), although both had the majority of songs in simple 4/4. At the moment, I can't think of a song in either band's repertoire played in 3/4.

But more importantly, the things which aren't similar:

1. Line-up stability: VdGG had an almost unchanging line-up whereas KC was more a "revolving door", partially through circumstance and partially through design
2. Guitar vs Organ, obviously
3. KC were almost always short of written material whereas VdGG always had a surfeit.
4. As a result, there are lengthy passages of improvisation scattered all around the KC catalogue whereas VdGG didn't release their improvisations (apart from the coda to 'Lemmings' and the lengthy instrumental part of 'Meurglys'. There are also the guitar led instrumentals ('Sailors tale', 'Larks tongue', 'Fracture', 'Red') which VdGG didn't need.
5. Vocalists! Whilst Greg Lake and John Wetton could match Hammill on the ballad side, none of them could "bare their tonsils to the world" in the same way that Hammill frequently did.

I think that the most important difference was that VdGG had the epic songs composed by Hammill and arranged by everyone. The actual songs as performed by KC tended to be simple affairs (the vocal sections of 21stCSM are the ultimate in simplicity) and the arrangements - at least, for three albums - were totally defined by one person. KC were much better when they played as a band (first album, fifth-seventh) than when they were part players (second-fourth albums).

The most VdGG-like song played by KC is probably 'Cirkus', but even that ranks unfavourably to 'Darkness'.

[SO: 3387; 2, 12, 30]

[MP&P: 356; 0, 0, 4]

Playing with fire (DCI Banks, TV)

I have now seen both episodes of 'Playing with fire'. I find it very difficult to comment on the programmes because the plot was very different from the book. I couldn't watch them as if the story were new and unknown to me.

Somehow I get the feeling that I wouldn't continue watching this series if I didn't know and love the books.

I could list the differences but there seem to be so many, and what's the point of doing so, anyway? It's not going to change anything.

On the other hand, there is the excellent series New Tricks, which I very much enjoy - but don't record for posterity. My wife even watched most of an episode yesterday - she had to go before the end, but asked me to tell her 'who dun it'. As it happens, that episode left the issue unresolved.

Learning curves

Yesterday I read the book "Learning curves (a novel of sex, suits and business affairs)" by Gemma Townley. The ride to and from Haifa bay was long enough for me to start and finish reading the book. I discovered the book because it fell into the category of "business novels", although this was somewhat different to other business novels which I have read. I have to point out at the very beginning that there was very little sex (certainly not graphic) and even less about suits in the novel.

The story was basically about three things

• Heroine Jen is sent to study for an MBA degree, so there's MBA material quoted
• Jen is to infiltrate into the organisation which is running the MBA and so find out about shady building deals in Indonesia
• Jen is also a woman and thus has to deal with affairs of the heart

The MBA part is rather unlike life (or unlike my MBA course): it does not take place under the auspices of a university; it's run by a consulting company. One wonders which academic credentials would accrue to such a 'degree'. The actual MBA material quoted is quite good, but it took me a while to identify from which of my courses it was coming. After a while, I realised that the characters were discussing in depth various subjects which arose from one of the analytical tools used in the marketing course. We covered this subject in one or two sessions whereas they seem to have spent all term on this. What about such mundane subjects as accounting, economics or finance? Ne'er a word.

The 'shady deals in Indonesia' plot was actually quite good, but apart from being mentioned at the beginning - the whole pretext for Jen going to study - and the denouement at the end, there isn't very much development, so the climax comes over as a deux ex machina ending. Shame.

As for the 'chic lit' plot which occupied no small amount of the book, I have read better. As far as I am concerned, most of this was extraneous and could easily have excised, but I am aware that this might be the genre which would attract most readers to this book.

It may just be a fault of the layout of the Kindle edition, but several times the point of view jumped from one character in one scene to another character in another scene. At times this was quite disturbing and even confusing. I call this poor editing.

To conclude: a good idea which went wrong in execution. Where was the editor?

Motorbike problems

About a week or so ago, I noticed that the speedometer on my motorbike wasn't working. This didn't bother me too much as my driving is not normally dependent on the bike's speed. I drive along two regular routes: to and from work (mainly along country lanes) and to and from the railway station. Only on the way back from the railway station is there a section when I can open the throttle, and by experience, 90 kph is more than fast enough.

The fuel gauge has never been accurate: when the tank is full, the gauge shows something like 120%, and when the gauge shows empty, I can still drive about another 50 km. So I don't base my refueling decisions on the fuel gauge but rather on kilometers driven: a full tank is good for 165-170 km. As I normally drive 10 km a day, refueling should be once every 17 days, which is slightly over three weeks (I don't drive the bike on Fridays and Saturdays).

It's been a few weeks since I last refueled, so I was thinking that the time had come. But when I checked the odometer, this showed that I had traveled only about 90 km, so there should be plenty of fuel left. I checked again on Sunday - still no need to refuel.

Yesterday, on the way to the railway station (and thus also needing to arrive at a specific time), the engine suddenly stopped. I imagined that the fuel tank was empty, so I pushed the bike to the nearby petrol station and refilled with 6.1 litres petrol (ie a full tank). I rushed off to the train station but saw the train leave just as I was arriving.

So I drove to work and checked the odometer, which was still showing about 90km more than the previous refuelling (I keep a spreadsheet of dates, kilometers and litres in order to work out the average consumption, when I need to refuel, etc). Slowly, I was reaching a sad conclusion: the odometer must have stopped working at or around the same time when the speedometer ceased working. I checked this hypothesis when I drove back to the railway station an hour later: the reading on the odometer hadn't changed.

I didn't think that this was possible: I know that many people who want to sell their cars would be only too pleased to disconnect the odometer so that the mileage shown is not the real mileage.

Anyway, the bike will be collected this evening and taken to the garage for repair. I don't need the bike tomorrow as

It is my daughter's wedding!