I'm afraid I couldn't begin to answer the first part of your question. Casualty figures of any type are always extremely difficult to narrow down and can never be specific. Records when available tell us only as much as whether a soldier was wounded, died of wounds or was killed in action. There's no way of knowing how exactly this took place without a personal account to tell us, and naturally these are extremely rare. Most men who died did so, it's sad to say, without anyone recording the precise circumstances. In certain cases we can make some limited assumptions--so if someone died in an assault, and we have accounts which mention particularly bad enemy machine gun fire, we can guess that the person in question was probably killed by a machine gun. But the extreme confusion which reigned in every battle means we can't possibly put a number on the amount of troops killed by their own side.

To attempt to answer your second question: what we can do is make some broad assumptions about how often this could happen, according to the accounts we have. For instance, the battle I'm currently researching, Aubers Ridge (9 May 1915), did see some considerable friendly fire. We know from personal accounts that German shelling was quite low-scale, but many shells fired by British guns fell short on the attacking troops, both during the initial bombardment while they waited to advance, and later after they had gone to ground in no-man's-land due to the failure of the assault. We also know that the cause of this was low quality ammunition and wear on the guns' barrels, which caused the driving bands of shells to fly off with negative effects on both accuracy and range. We can be fairly sure, therefore, that the artillery was registered (aimed) effectively, since the problem arose from mechanical defects. Firing well-maintained guns at pre-planned targets, the artillery could usually avoid hitting their own side! We know the British lost a lot of men before the assault even began, but since the Germans were also pouring rifle and machine-gun fire into the attacking troops at the same time, it's very difficult to estimate whether enemy action or friendly artillery caused the more casualties. It's fair to assume that such events decreased in frequency as British industry began to provide more and better quality guns and ammunition.

We can however assume that in cases where assaults succeeded, casualties from friendly fire were higher. With no way to communicate directly with friendly artillery, troops caught in a friendly bombardment had little option but to weather it. This happened in several places on the first day of the Somme, and was reasonably common throughout 1916-17. This tended to take place on occasions where attacking troops outran their supporting artillery and ended up in areas which were assumed to still be occupied by enemy troops--particularly once the British army began to use more advanced tactics like the creeping barrage, but before these tactics were fully perfected. Given the protection provided by trenches--despite their reputation, they were undoubtedly the safest places to be if a battle was raging--casualties still may not have been unduly high--the effect of artillery on entrenched troops, whether they were friend or foe, was usually to paralyze rather than decimate.

Friendly fire from rifles and machine gun fire is again virtually impossible to estimate--how would one determine whether a casualty was hit by a friendly or enemy bullet? This would be technically possible with a huge amount of detective work, but nobody had time to do this and such details have been lost to history in all but a very few cases. Undoubtedly this took place, but I don't know of any large scale cases (although I'd love to hear from anyone who does). I wouldn't overall say friendly fire was ever particularly widespread or had a serious effect, any more than one would expect in any war, but this is based upon by own very limited knowledge of personal accounts, because there are no concrete statistics to draw upon.

I think 'unintentional suicide' was even rarer. As far as I know there was little chance of a pilot shooting himself down. Synchronization gears limited the rate of fire which machine guns could use, since a higher rate of fire put more strain on the mechanism and increased the chance of a failure. But only those mechanisms which were consistently reliable were employed in large numbers at the front. It wasn't worth risking planes and pilots by using unreliable mechanisms.

Tanks didn't tend to blow up unless they were hit. They did, however, have a propensity to kill their crews due to the conditions inside. It was always extremely loud and hot, and the unshielded engines tended to release exhaust fumes into the tank. All these factors meant that a crew could perform for little more than an hour or so before their effectiveness plummeted, but of course battle required them to operate for far longer, and in many cases this proved fatal.

There is another case which arguably might be called friendly fire. Many thousands of Chinese labourers, brought in to help clear up the battlefields after the war, were killed by unexploded ordnance, much of it, of course, fired from Allied guns.

I hope this is helpful, and apologise that I can't be more specific, but I know of no proper assessment of friendly fire casualties in the First World War and my own knowledge is limited.